Data detection method and system based on gravity, magnetic, electric and seismic signal joint gridding

Abstract

The invention discloses a data detection method and system based on gravitational-magnetic-electric-seismic signal joint gridding, and relates to the technical field of geological prospecting, and the method comprises the steps: determining a research region according to an exploration target and geological conditions, and dividing the research region into grids with the same size; on each grid node, gravity, magnetic, electric and seismic signals are collected; the collected gravity, magnetic and electric shock signals are preprocessed; performing feature extraction on the preprocessed gravity-magnetic-electric-seismic signals by using an independent component analysis algorithm, and performing data fusion by using a convolutional neural network in deep learning; establishing an initial value estimation incidence relation model, and obtaining an initial value prediction result based on the fused data; according to the initial value prediction result, geological parameters are obtained through solving by using a joint inversion algorithm. According to the invention, efficient combined acquisition of gravity, magnetic, electric and seismic signals is realized, the precision and quality of acquired data are improved, and unified management and fusion analysis of the data are realized.

Description

Technical Field

[0001] The present invention relates to the field of geological exploration technology, and more particularly to a data detection method and system based on combined gridding of gravity, magnetic, electric and seismic signals. Background Art

[0002] Currently, in geological exploration, gravity, magnetic, electric, and seismic signals are often collected independently using a single signal. Even when combined, these signals often suffer from poor equipment coordination, low data collection efficiency, and low accuracy. For example, traditional gravity signal acquisition equipment is susceptible to environmental interference in areas with complex terrain, leading to large measurement errors. During magnetic signal acquisition, the instrument's inherent limitations hinder its ability to capture weak magnetic field signals. When acquiring electrical signals, the electrode layout and position significantly impact the results, leading to long data acquisition cycles. Seismic signal acquisition places high demands on the acquisition equipment's response speed and data processing capabilities. Existing equipment often struggles to accurately collect and process seismic wave signals under complex geological conditions.

[0003] At the same time, the lack of a unified acquisition plan and the irrational distribution of acquisition points during the acquisition process prevented the collected data from fully and evenly covering the target area, making it difficult to construct a high-precision underground geological model and hindering subsequent analysis and judgment of geological structure and resource distribution. Furthermore, the data transmission and storage formats between different types of signal acquisition equipment were not standardized, making data fusion difficult and preventing the full utilization of the advantages of combined acquisition of gravity, magnetic, electric, and seismic signals. As geological exploration gradually expands into deeper and more complex areas, the requirements for the accuracy and completeness of underground information are becoming increasingly stringent, and traditional acquisition methods and equipment are no longer able to meet actual needs.

[0004] Therefore, how to provide a data detection method and system based on the joint gridding of gravity, magnetic, electric and seismic signals to achieve efficient joint acquisition of gravity, magnetic, electric and seismic signals, improve the accuracy and quality of acquired data, and realize unified management and fusion analysis of data is an urgent problem that technical personnel in this field need to solve. Summary of the Invention

[0005] In view of this, the present invention provides a data detection method and system based on the joint gridding of gravity, magnetic, electric and seismic signals, which realizes the efficient joint acquisition of gravity, magnetic, electric and seismic signals, improves the accuracy and quality of the acquired data, optimizes the layout of the acquisition points, and realizes the unified management and fusion analysis of the data.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a data detection method based on combined gridding of gravity, magnetic, electric and seismic signals, comprising:

[0007] Determine a study area based on exploration objectives and geological conditions, and divide the study area into grids of equal size;

[0008] At each grid node, gravity, magnetic, electric and seismic signals are collected;

[0009] Pre-process the collected gravity, magnetic, electric and seismic signals;

[0010] The independent component analysis algorithm is used to extract features from the pre-processed gravity, magnetic, electric and seismic signals, and the convolutional neural network in deep learning is used for data fusion;

[0011] Establish an initial value estimation correlation model and obtain the initial value prediction results based on the fused data;

[0012] According to the initial prediction results, the geological parameters are solved using the joint inversion algorithm.

[0013] Preferably, the independent component analysis algorithm is used to extract features from the pre-processed gravity, magnetic, electric and seismic signals, including:

[0014] The pre-processed gravity, magnetic, electric and seismic signals are combined into a mixed signal matrix;

[0015] Perform whitening on the mixed signal matrix X to obtain the whitened data matrix Z;

[0016] Initialize a separation matrix W and iteratively update the separation matrix W to maximize the non-Gaussianity of the separated signal Y=WZ. The obtained separated signal Y is the characteristic of gravity, magnetic, electric and seismic data.

[0017] Preferably, based on the acquired gravity, magnetic, electric and seismic data features, a convolutional neural network in deep learning is used to perform data fusion, including:

[0018] In the convolution layer, convolution kernels of different sizes and strides are used to perform sliding convolution on the gravity, magnetic, electric and seismic data features to extract local features of the data;

[0019] The pooling layer is used to downsample the features extracted by the convolutional layer;

[0020] The fully connected layer connects the feature vectors after convolution and pooling, maps them to the final output space, and outputs the attribute data of gravity, magneto-electro-seismic signals.

[0021] Preferably, an initial value estimation correlation model is established to obtain an initial value prediction result based on the fused data, including:

[0022] The fused gravity, magnetic, electric and seismic signals are input into the initial value estimation correlation model. The initial value estimation correlation model analyzes and processes the input data according to the learned correlation relationship and outputs the initial value prediction results required for the gravity, magnetic, electric and seismic joint inversion.

[0023] Preferably, the expression of the joint inversion algorithm is as follows:

[0024] S=S data +λS model ;

[0025] Among them, S data represents the data fitting term, S model Represents the model constraint term; λ is the regularization parameter, which is used to balance the weights of the data fitting term and the model constraint term;

[0026] The matching degree between the inversion results and the observed data is measured by the data fitting term, and the smoothing constraint algorithm is used to constrain the model.

[0027] Preferably, the degree of matching between the inversion results and the observed data is measured by the data fitting term, including:

[0028] For earthquake data d sobs and earthquake theory calculation data d scal , gravity data d gobs and gravity theory calculation data d gcal , magnetic data d mobs and magnetic theory calculation data d mcal , electric field data d eobs and electric field theory calculation data d ecal , the data fitting term is expressed as:

[0029]

[0030] Among them, n s 、n g 、n m and n e are the number of seismic, gravity, magnetic and electric field data respectively.

[0031] Preferably, a smooth constraint algorithm is used to constrain the model, including:

[0032] The smoothness constraint is expressed as:

[0033]

[0034] Among them, m i is the i-th parameter of the geological model, and N is the total number of parameters.

[0035] Model constraints are used to introduce geological prior information to ensure the rationality of the inversion results. Smoothness constraints require that model parameters vary as smoothly as possible in space to avoid drastic jumps.

[0036] Preferably, a data detection system based on combined gridding of gravity, magnetic, electric and seismic signals comprises:

[0037] A grid division module is used to determine a research area according to exploration objectives and geological conditions, and divide the research area into grids of the same size;

[0038] Signal acquisition module, used to collect gravity, magnetic, electric and seismic signals at each grid node;

[0039] Signal preprocessing module, used to preprocess the collected gravity, magnetic, electric and seismic signals;

[0040] The data fusion module is used to extract features from the pre-processed gravity, magnetic, electric and seismic signals using the independent component analysis algorithm and perform data fusion using the convolutional neural network in deep learning;

[0041] The initial value prediction module is used to establish an initial value estimation correlation model and obtain the initial value prediction results based on the fused data;

[0042] The geological parameter solution module is used to obtain geological parameters based on the initial value prediction results using the joint inversion algorithm.

[0043] Through the above technical solution, it can be seen that compared with the existing technology, the present invention discloses a data detection method and system based on the joint gridding of gravity, magnetic, electric and seismic signals, including: determining the study area according to the exploration target and geological conditions, and dividing the study area into grids of equal size; collecting gravity, magnetic, electric and seismic signals at each grid node; preprocessing the collected gravity, magnetic, electric and seismic signals; using the independent component analysis algorithm to extract features from the preprocessed gravity, magnetic, electric and seismic signals, and using the convolutional neural network in deep learning to perform data fusion; establishing an initial value estimation correlation relationship model, and obtaining initial value prediction results based on the fused data; and using the joint inversion algorithm to solve and obtain geological parameters based on the initial value prediction results. The present invention realizes the efficient joint collection of gravity, magnetic, electric and seismic signals, improves the accuracy and quality of the collected data, and realizes the unified management and fusion analysis of data. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0045] Figure 1 A flow chart of a data detection method based on combined gridding of gravity, magnetic, electric and seismic signals provided by the present invention.

[0046] Figure 2This is a structural schematic diagram of a data detection system based on the combined gridding of gravity, magnetic, electric and seismic signals provided by the present invention. DETAILED DESCRIPTION

[0047] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0048] The embodiment of the present invention discloses a data detection method based on the combined gridding of gravity, magnetic, electric and seismic signals. Figure 1 Shown, including:

[0049] Determine the study area based on the exploration objectives and geological conditions, and divide the study area into grids of equal size using a regular grid division method; the grid side length is determined based on the exploration accuracy requirements and the geological characteristics of the study area, and is generally in the range of [X1, X2] meters to ensure that each grid can contain sufficient geological information and meet the accuracy requirements of data processing;

[0050] At each grid node, gravity, magnetic, electric and seismic signals are collected;

[0051] Specifically, high-precision gravimeters, magnetometers, electromagnets, and seismographs are selected to ensure accurate acquisition of gravity, magnetic, electric, and seismic signals. Within the target exploration area, a grid layout is used to divide the area into multiple square or rectangular grids based on the geological structure and exploration needs. Sensors are deployed at each grid node and key locations, allowing different sensor types to work together to achieve comprehensive coverage and monitoring of the area. In areas with complex geology, sensors are appropriately encrypted to ensure the integrity and accuracy of data collection.

[0052] A high-precision atomic clock is used as the time reference, and a synchronous clock signal is sent to each sensor through a wireless or wired network to ensure that all sensors start data collection at the same time point, the collection frequency remains consistent, and data synchronization is achieved.

[0053] Preprocess the collected gravity, magnetic, electric, and seismic signals; apply filtering algorithms to remove noise interference. For example, using wavelet filtering, select an appropriate wavelet basis and number of decomposition levels based on the frequency characteristics of the noise and signal to effectively filter out high-frequency noise and low-frequency interference signals, while retaining the useful signal. Standardize the data to unify data with different dimensions and numerical ranges collected by different sensors to the same scale. Use normalization methods to make the data comparable in subsequent analysis and improve processing accuracy. Use interpolation algorithms to supplement missing data. Based on the spatial distribution characteristics of the data, select an appropriate interpolation method, such as Kriging interpolation, and combine the spatial correlation of surrounding data to accurately estimate missing values and ensure data integrity.

[0054] The independent component analysis (ICA) algorithm was used to extract features from the preprocessed gravity, magnetic, electric, and seismic signals. ICA can separate independent components from mixed signals, extracting key features reflecting geological structure and physical properties from complex data, providing effective data support for subsequent analysis. Data fusion was also performed using convolutional neural networks (CNNs), a deep learning framework. CNNs have powerful feature learning capabilities and can automatically learn deep-level features of data. By constructing a CNN model, the features of the gravity, magnetic, electric, and seismic data were integrated, potential connections between the data were discovered, and the accuracy of geological parameter extraction was improved.

[0055] Establish an initial value estimation correlation model and obtain the initial value prediction results based on the fused data;

[0056] According to the initial prediction results, the geological parameters are solved using the joint inversion algorithm.

[0057] Based on seismic attribute data such as wave impedance and velocity, a correlation relationship with gravity, magneto-electric attribute data is established through methods such as neural networks to provide accurate initial values for inversion.

[0058] Based on the established initial values, a joint inversion algorithm is used to solve the geological parameters: a gravity-magnetic-electrical-seismic joint inversion method is adopted, and the complementarity of different geophysical methods is considered. The resistivity, polarizability, magnetic susceptibility, wave velocity and other parameters are obtained through comprehensive inversion to achieve accurate imaging of the underground geological structure.

[0059] Specifically, the independent component analysis algorithm is used to extract features from the pre-processed gravity, magnetic, electric and seismic signals, including:

[0060] The preprocessed gravity, magnetic, electric and seismic signals are organized into a mixed signal matrix. The preprocessed gravity, magnetic, electric and seismic data are organized according to specific rules to form a format suitable for ICA processing. The gravity, magnetic, electric and seismic data are defined as M (gravity data), G (magnetic data), E (electrical data), and S (seismic data). These data are aligned by time series or spatial position to form a mixed signal matrix X, where each column represents the measurement value of a sensor at a different time or position. The ICA algorithm estimates the original independent components through an iterative algorithm based on the mutual independence of the independent components in the mixed signal and the non-Gaussian nature of the signal itself.

[0061] Whiten the mixed signal matrix X to remove the correlation in the data and obtain the whitened data matrix Z. This step is to make subsequent calculations more efficient and stable and reduce redundant information between data.

[0062] Initialize a separation matrix W. The initial values of this matrix are typically random, but must meet certain conditions to ensure convergence. Iteratively update the separation matrix W to maximize the non-Gaussianity of the separated signal Y = WZ. The resulting separated signal Y represents the characteristics of the gravity, magnetic, electric, and seismic data. During the iterations, gradient ascent or other optimization algorithms are used to adjust the elements of the separation matrix W, continuously approaching the optimal solution. Non-Gaussianity measures such as negative entropy and kurtosis are used.

[0063] After multiple iterations, when the separation matrix W converges to a certain degree, the resulting separation signal Y is the estimated independent component. These independent components contain key features reflecting geological structure and physical properties. Some independent components may correspond to information such as magnetic differences, electrical differences, or density differences between different underground geological bodies.

[0064] Specifically, based on the characteristics of the acquired gravity, magnetic, electric and seismic data, a convolutional neural network in deep learning is used for data fusion, including:

[0065] In the convolutional layer, convolution kernels of varying sizes and strides are used to perform sliding convolutions on the gravity, magnetic, electric, and seismic data to extract local features. For two-dimensional seismic data maps, the convolution kernels can slide horizontally and vertically to extract geological structural features at different scales. The convolutional layer parameters (such as kernel size, number, and stride) are adjusted based on data characteristics and experimental results to achieve optimal feature extraction.

[0066] The pooling layer downsamples the features extracted by the convolutional layer, reducing the dimensionality of the data while retaining important feature information. Common pooling methods include max pooling and average pooling. The appropriate pooling method can be selected based on the characteristics of the data. Max pooling selects the maximum value within a local area as the pooling result, which can highlight key features of the data. Average pooling calculates the average value within a local area, which is effective for smoothing data.

[0067] The fully connected layer connects the feature vectors after convolution and pooling and maps them to the final output space, outputting the attribute data of the gravity, magnetic, electric and seismic signals. In the fully connected layer, the features are weighted and summed using the weight matrix and bias term to achieve data classification or regression tasks.

[0068] The embodiment of the present invention uses a large number of gravity, magnetic, electric and seismic data samples to train the constructed CNN model. These samples contain data under different geological conditions to ensure that the model can learn a wide range of data features and patterns.

[0069] The training data is labeled with real geological parameter information, such as the actual values of underground geological parameters such as density, magnetism, and electrical properties. This labeled information will serve as a supervisory signal for model training, guiding the model to learn the mapping relationship between data and geological parameters.

[0070] A loss function is used to measure the difference between the model's predictions and the true annotations. For example, mean squared error (MSE) is used for regression tasks, and cross-entropy loss is used for classification tasks. An appropriate optimizer (such as stochastic gradient descent (SGD), Adagrad, or Adadelta) is selected to adjust the model's parameters to minimize the loss function. During training, parameters such as the optimizer's learning rate are adjusted based on the model's training progress and convergence rate to improve training results.

[0071] The gravity, magnetic, electric, and seismic data features are fed into the CNN model. The model's predictions are calculated through forward propagation. The error between the predictions and the ground truth annotations is then calculated using a loss function. Next, the gradient of the error with respect to the model parameters is calculated using a backpropagation algorithm, and the optimizer updates the model parameters based on the gradient. This process is repeated until the model's loss function on the training set converges to a small value, indicating that the model has learned the characteristics and patterns of the data.

[0072] The trained CNN model is applied to actual gravity, magnetic, electric, and seismic data fusion and geological parameter extraction. The gravity, magnetic, electric, and seismic signal features extracted by the ICA algorithm are input into the trained CNN model, which then outputs fused geological parameter results. These results can be used to draw geological profiles and determine the location and shape of geological bodies, providing strong support for geological exploration and research. In practical applications, the model can be further optimized and adjusted as needed to adapt to different geological conditions and exploration tasks.

[0073] Model performance was evaluated using metrics such as mean squared error (MSE) and mean absolute error (MAE). MSE, which averages the sum of the squares of the errors between the predicted and true values, reflects the overall degree of error in the model's predictions. MAE, which averages the absolute errors between the predicted and true values, is more sensitive to outliers. The neural network structure and parameters were adjusted based on the evaluation results.

[0074] Specifically, an initial value estimation correlation model is established, and the initial value prediction results are obtained based on the fused data, including:

[0075] The fused gravity, magnetic, electric, and seismic signals are fed into the initial value estimation correlation model. This model analyzes and processes the input data based on the learned correlations and outputs the initial value predictions required for the gravity, magnetic, electric, and seismic joint inversion. These initial value predictions include initial estimates of resistivity, polarizability, magnetic susceptibility, wave velocity, and other parameters, providing accurate and reliable initial parameters for subsequent joint inversion.

[0076] Specifically, the expression of the joint inversion algorithm is as follows:

[0077] S=S data +λS model ;

[0078] Among them, S data represents the data fitting term, S model Represents the model constraint term; λ is the regularization parameter, which is used to balance the weights of the data fitting term and the model constraint term;

[0079] The matching degree between the inversion results and the observed data is measured by the data fitting term, and the smoothing constraint algorithm is used to constrain the model.

[0080] Specifically, the degree of matching between the inversion results and the observed data is measured by the data fitting term, including:

[0081] For earthquake data d sobs and earthquake theory calculation data d scal , gravity data d gobs and gravity theory calculation data d gcal , magnetic data d mobsand magnetic theory calculation data d mcal , electric field data d eobs and electric field theory calculation data d ecal , the data fitting term is expressed as:

[0082]

[0083] Among them, n s 、n g 、n m and n e are the number of seismic, gravity, magnetic, and electric field data, respectively;

[0084] Specifically, a smooth constraint algorithm is used to constrain the model, including:

[0085] The smoothness constraint is expressed as:

[0086]

[0087] Among them, m i is the i-th parameter of the geological model, and N is the total number of parameters.

[0088] Model constraints are used to introduce geological prior information to ensure the rationality of the inversion results. Smoothness constraints require that model parameters vary as smoothly as possible in space to avoid drastic jumps.

[0089] In a specific embodiment of the present invention, a data detection system based on combined gridding of gravity, magnetic, electric and seismic signals is provided. Figure 2 Shown, including:

[0090] A grid division module is used to determine a research area according to exploration objectives and geological conditions, and divide the research area into grids of the same size;

[0091] Signal acquisition module, used to collect gravity, magnetic, electric and seismic signals at each grid node;

[0092] Signal preprocessing module, used to preprocess the collected gravity, magnetic, electric and seismic signals;

[0093] The data fusion module is used to extract features from the pre-processed gravity, magnetic, electric and seismic signals using the independent component analysis algorithm and perform data fusion using the convolutional neural network in deep learning;

[0094] The initial value prediction module is used to establish an initial value estimation correlation model and obtain the initial value prediction results based on the fused data;

[0095] The geological parameter solution module is used to obtain geological parameters based on the initial value prediction results using the joint inversion algorithm.

[0096] In one specific embodiment of the present invention, consider a metal mining exploration area with complex geological conditions, multiple rock layers, and potential deposits of copper, lead, zinc, and other metal minerals. The goal of this exploration is to accurately determine the underground mineral distribution and geological structure to provide critical data support for subsequent mining.

[0097] Gridding: The scope of the study area is determined based on the exploration accuracy requirements and regional geological characteristics. A regular grid is used to divide the area into squares with sides of 150 meters. This grid size ensures sufficient geological information is obtained while meeting the accuracy requirements of data processing.

[0098] Signal Acquisition: High-precision gravimeters (such as the CG-5 gravimeter), magnetometers (G856 proton magnetometers), electrical instruments (GDP-32II multifunctional electrical instruments), and seismometers (ES-1221F seismometers) are used. Based on the geological structure, sensors are deployed at each grid node and key locations where geological anomalies may exist (such as near fault zones). These different sensor types work together to achieve comprehensive regional monitoring. Using a high-precision atomic clock as the time reference, synchronized clock signals are sent to each sensor via a wireless transmission network, ensuring that all sensors initiate data acquisition at the same time. The acquisition frequency is uniformly set to 100 Hz to ensure data synchronization.

[0099] Signal preprocessing: Wavelet filtering algorithms were used to remove noise from the collected signals. Based on the frequency characteristics of the signal and noise, the db6 wavelet basis was selected, and the decomposition layer was set to 6, effectively filtering out noise while retaining the useful signal. Normalization was used to standardize the data, unifying data with different dimensions and numerical ranges collected by different sensors to a scale of [0, 1] to enhance data comparability. Missing data was supplemented using the Kriging interpolation algorithm, combined with the spatial correlation of surrounding data, to ensure data integrity.

[0100] Data fusion and model building: Preprocessed gravity, magnetic, and seismic data were spatially aligned to form a mixed signal matrix X, which was then whitened to obtain matrix Z. The separation matrix W was initialized and iteratively updated. Kurtosis was used as a measure of non-Gaussianity, and the gradient ascent method was used to maximize the non-Gaussianity of the separated signal Y = WZ to obtain the gravity, magnetic, and seismic data features. A CNN model was constructed, consisting of four convolutional layers, three pooling layers, and two fully connected layers. In the convolutional layers, different kernel sizes (3×3, 5×5) and strides (1, 2) were set to extract geological structural features at different scales. The pooling layers used max pooling to retain key features. The fully connected layers output the gravity, magnetic, and seismic signal attribute data. A large number of gravity, magnetic, and seismic data samples under different geological conditions were collected to train the model. The real geological parameters were annotated. The mean squared error (MSE) was used as the loss function, and the Adadelta optimizer was used to adjust the model parameters. The model was trained repeatedly until the loss function converged.

[0101] Initial value prediction and geological parameter solution: The fused gravity, magnetic, electric, and seismic signals are input into the initial value estimation correlation model to obtain the initial value prediction results required for the joint gravity, magnetic, electric, and seismic inversion, such as initial estimates of resistivity, polarizability, magnetic susceptibility, and wave velocity. Using the joint inversion algorithm, an appropriate regularization parameter λ is set, and calculations are performed in conjunction with data fitting terms and model constraints. Ultimately, through multiple iterative calculations, a comprehensive inversion is performed to obtain parameters such as resistivity, polarizability, magnetic susceptibility, and wave velocity of the underground geological structure. This allows the creation of an accurate geological profile, clearly depicting the underground geological structure and potential mineral distribution, providing a strong basis for subsequent mineral mining planning.

[0102] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0103] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data detection method based on the combined gridding of gravity, magnetic, electric and seismic signals, characterized in that: include: Determine a study area based on exploration objectives and geological conditions, and divide the study area into grids of equal size; At each grid node, gravity, magnetic, electric and seismic signals are collected; Pre-process the collected gravity, magnetic, electric and seismic signals; The independent component analysis algorithm is used to extract features from the pre-processed gravity, magnetic, electric and seismic signals, and the convolutional neural network in deep learning is used for data fusion; Establish an initial value estimation correlation model and obtain the initial value prediction results based on the fused data; According to the initial prediction results, the geological parameters are solved using the joint inversion algorithm.

2. The data detection method based on combined gridding of gravity, magnetic, electric and seismic signals according to claim 1 is characterized in that: The independent component analysis algorithm is used to extract features from the pre-processed gravity, magnetic, electric and seismic signals, including: The pre-processed gravity, magnetic, electric and seismic signals are combined into a mixed signal matrix; Perform whitening on the mixed signal matrix X to obtain the whitened data matrix Z; Initialize a separation matrix W and iteratively update the separation matrix W to maximize the non-Gaussianity of the separated signal Y=WZ. The obtained separated signal Y is the characteristic of gravity, magnetic, electric and seismic data.

3. The data detection method based on combined gridding of gravity, magnetic, electric and seismic signals according to claim 2 is characterized in that: Based on the characteristics of the acquired gravity, magnetic, electric and seismic data, a convolutional neural network in deep learning is used for data fusion, including: In the convolution layer, convolution kernels of different sizes and strides are used to perform sliding convolution on the gravity, magnetic, electric and seismic data features to extract local features of the data; The pooling layer is used to downsample the features extracted by the convolutional layer; The fully connected layer connects the feature vectors after convolution and pooling, maps them to the final output space, and outputs the attribute data of gravity, magneto-electro-seismic signals.

4. The data detection method based on combined gridding of gravity, magnetic, electric and seismic signals according to claim 1 is characterized in that: Establish an initial value estimation correlation model and obtain initial value prediction results based on the fused data, including: The fused gravity, magnetic, electric and seismic signals are input into the initial value estimation correlation model. The initial value estimation correlation model analyzes and processes the input data according to the learned correlation relationship and outputs the initial value prediction results required for the gravity, magnetic, electric and seismic joint inversion.

5. The data detection method based on combined gridding of gravity, magnetic, electric and seismic signals according to claim 1 is characterized in that: The expression of the joint inversion algorithm is as follows: S=S data +λS model ; Among them, S data represents the data fitting term, S model Represents the model constraint term; λ is the regularization parameter, which is used to balance the weights of the data fitting term and the model constraint term; The matching degree between the inversion results and the observed data is measured by the data fitting term, and the smoothing constraint algorithm is used to constrain the model.

6. The data detection method based on combined gridding of gravity, magnetic, electric and seismic signals according to claim 5 is characterized in that: The degree of matching between the inversion results and the observed data is measured by the data fitting term, including: For earthquake data d sobs and earthquake theory calculation data d scal , gravity data d gobs and gravity theory calculation data d gcal , magnetic data d mobs and magnetic theory calculation data d mcal , electric field data d eobs and electric field theory calculation data d ecal , the data fitting term is expressed as: Among them, n s 、n g 、n m and n e are the number of seismic, gravity, magnetic and electric field data respectively.

7. The data detection method based on combined gridding of gravity, magnetic, electric and seismic signals according to claim 5 is characterized in that: The smooth constraint algorithm is used to constrain the model, including: The smoothness constraint is expressed as: Among them, m i is the i-th parameter of the geological model, and N is the total number of parameters; Model constraints are used to introduce geological prior information to ensure the rationality of the inversion results.

8. A data detection system based on the combined gridding of gravity, magnetic, electric and seismic signals, applying the data detection method based on the combined gridding of gravity, magnetic, electric and seismic signals according to any one of claims 1 to 7, characterized in that: include: A grid division module is used to determine a research area according to exploration objectives and geological conditions, and divide the research area into grids of the same size; Signal acquisition module, used to collect gravity, magnetic, electric and seismic signals at each grid node; Signal preprocessing module, used to preprocess the collected gravity, magnetic, electric and seismic signals; The data fusion module is used to extract features from the pre-processed gravity, magnetic, electric and seismic signals using the independent component analysis algorithm and perform data fusion using the convolutional neural network in deep learning; The initial value prediction module is used to establish an initial value estimation correlation model and obtain the initial value prediction results based on the fused data; The geological parameter solution module is used to obtain geological parameters based on the initial value prediction results using the joint inversion algorithm.

Images