Active source magnetization-time domain electromagnetic induction polarization multi-parameter extraction method
By extracting magnetic susceptibility during the flat-top stage of the emission current using the active source magnetization-time-domain electromagnetic method, and combining it with the deep Q-network method to extract resistivity and polarizability during the turn-off stage, the problem of distinguishing between induced magnetization and residual magnetization is solved. This achieves high-precision extraction of multiple parameters, reduces the problem of multiple solutions, and improves the detection effect.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JILIN UNIVERSITY
- Filing Date
- 2025-03-27
- Publication Date
- 2026-07-21
AI Technical Summary
Existing technologies struggle to effectively distinguish and extract induced magnetization and residual magnetization in the Earth's magnetic field, leading to severe interference in magnetic susceptibility extraction. Furthermore, time-domain electromagnetic methods present challenges in extracting multiple parameters and multiple solutions in polymetallic mineral exploration.
The active source magnetization-temporal electromagnetic method is adopted. In the flat-top stage of the emission current, the Occam's method is used to extract the magnetic susceptibility. In the turn-off stage, the deep Q-network method is used in combination with the magnetic susceptibility constraint to extract the resistivity and polarizability. Multi-parameter extraction is performed through data segmentation and deep learning.
It improves the extraction accuracy of magnetic susceptibility, resistivity and polarizability, reduces the problem of multiple solutions, and improves the detection resolution and signal-to-noise ratio, making it suitable for the detection of complex multi-parameter underground media.
Smart Images
Figure CN120103499B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method for extracting multiple parameters of active source magnetization and time-domain electromagnetic induction polarization in the field of geophysical exploration, and particularly to a method for extracting magnetic susceptibility using the flat-top stage of a trapezoidal emission waveform and extracting resistivity and polarizability using the turn-off stage. Background Technology
[0002] In geophysical exploration, magnetic susceptibility is mostly extracted from static magnetic data. The total magnetic field strength is typically measured as the vector sum of the Earth's main magnetic field and the induced and remanent magnetization of magnetic anomalies. However, traditional static magnetic methods struggle to distinguish between induced and remanent magnetization, with remanent magnetization severely interfering with susceptibility extraction. Furthermore, the vertical component of the Earth's magnetic field is extremely weak, even near zero, in the magnetic equator region, resulting in anomalies in the induced magnetization data, making it difficult to estimate the number and susceptibility of magnetic anomalies. Active magnetic methods effectively compensate for the shortcomings of static magnetic methods by utilizing magnetic field signals excited by active sources to detect subsurface magnetic media, thus improving detection resolution and signal-to-noise ratio.
[0003] In polymetallic mineral exploration, time-domain electromagnetic methods (TDEM) exhibit inductive responses caused by conductivity, polarization responses caused by rechargeability, and magnetization responses caused by complex magnetic properties. Single-parameter extraction methods cannot extract all these parameters. Therefore, a method for simultaneous extraction of multiple parameters (resistivity, polarizability, and magnetic susceptibility) is urgently needed. Furthermore, using only TDEM data for multi-parameter extraction of resistivity, polarizability, and magnetic susceptibility is more prone to multiple-solution problems. Thus, how to achieve simultaneous and high-precision extraction of resistivity, polarizability, and magnetic susceptibility is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0004] The technical problem to be solved by this invention is to provide a method for extracting multiple parameters of active source magnetization and time-domain electromagnetic induction polarization. During the flat-top phase of the emission current, the magnetic susceptibility is extracted using the Occam's method based on active magnetization. During the off-phase phase of the emission current, resistivity and polarization are extracted using a deep Q-network method based on time-domain electromagnetic induction, with the extracted magnetic susceptibility as a constraint.
[0005] This invention is implemented as follows: A method for extracting multiple parameters of induced polarization from active source magnetization-time-domain electromagnetic detection includes:
[0006] 1) Separate the data for the flat-top phase and the shutdown time;
[0007] 2) The magnetic susceptibility parameter was extracted from the flat-top stage data using the active magnetic method;
[0008] 3) The extracted magnetic susceptibility is used as a constraint condition for time-domain electromagnetic method to extract resistivity and polarizability parameters;
[0009] 4) Obtain resistivity, polarizability and magnetic susceptibility parameters.
[0010] Further, in step 2, a parameter extraction model f = Gm is constructed, where m is the magnetic susceptibility, f is the response data, and G is the numerical simulation model operator; this model is then substituted into the Occam algorithm. Where m ref Let C be the magnetic susceptibility of the reference model, C be the roughness matrix composed of second-order differential operators, and W be the magnetic susceptibility of the reference model. z Let d be a diagonal matrix, D be the diagonal weighted matrix of the reference model, and d be the diagonal matrix. obs These are the observed data, F(m) is the predicted response of the magnetic susceptibility model, and λ and μ are trade-off parameters; the Occam parameter extraction model update formula is...
[0011] The magnetic susceptibility χ is extracted through iterative looping.
[0012] Furthermore, in step 3, the extracted magnetic susceptibility χ is used as a constraint condition to correct the response curve in the time-domain electromagnetic numerical simulation; a DQN parameter extraction model is constructed:
[0013] First, construct the objective function.
[0014]
[0015] Where (σ,η) represents the resistivity σ and polarizability η in the current state, (σ′,η′) represents the resistivity σ′ and polarizability η′ in the next state, (Δσ,Δη) represents the current action perturbation value, (Δσ′,Δη′) represents the next action perturbation value, and Q′ represents the updated state-action value. The action reward r is obtained in state (σ,η) and evaluated. γ is a discount factor; a larger γ emphasizes long-term rewards, while a smaller γ emphasizes short-term gains. θ is a neural network parameter. The target state-action value function y′=r+γmaxQ[(σ,η),(Δσ,Δη),θ] is constructed, where y′ represents the target Q value. The loss function L(θ)=E{{y′-Q[(σ,η),(Δσ,Δη),θ]} is constructed. 2}, where θ is the weight parameter trained in the neural network structure model.
[0016] Secondly, a Q-value network is constructed. DQN evaluates the current state-action value function through a target network and an estimation network. The target function obtains the target Q-value based on the neural network, and the target Q-value is used to estimate the Q-value at the next time step. The estimation network updates the network weights Δθ using stochastic gradient descent, where... It is the gradient operator, responsible for adjusting the parameters along the negative gradient direction. The gradient descent algorithm is...
[0017]
[0018] Finally, an experience replay pool is constructed. When the agent interacts with the environment, a sample database can be obtained. The sample database is stored in the established experience pool. A portion of the data is randomly selected from the experience pool for training samples, and then the training samples are fed into the neural network for training.
[0019] Compared with existing technologies, the advantages of this invention are as follows: For the problem of extracting multiple parameters (resistivity, polarizability, and magnetic susceptibility) from polymetallic ores, the data is segmented, and magnetic susceptibility is extracted during the flat-top stage, providing constraints for parameter extraction during the turn-off stage, reducing the difficulty of parameter extraction and mitigating the problem of multiple solutions. During the flat-top stage of the emission current, there is a linear relationship between the magnetic susceptibility parameter and the response of the active magnetic method; using the Occam's algorithm for parameter extraction can effectively improve the efficiency of parameter extraction. During the turn-off stage of the emission current, magnetic susceptibility is used as a constraint, reducing the difficulty of parameter extraction. A deep Q-network method is used for parameter extraction, which integrates the perceptual capabilities of deep learning and the decision-making capabilities of reinforcement learning. It does not require building a large dataset and is suitable for handling complex multi-effect, multi-parameter extraction problems. Attached Figure Description
[0020] Figure 1 This is a schematic diagram of the active source magnetization-time-domain electromagnetic induction polarization multi-parameter extraction method;
[0021] Figure 2 This is a flowchart of magnetization parameter extraction based on Occam's algorithm;
[0022] Figure 3 This is a flowchart of the induction-polarization parameter extraction based on the deep Q-network algorithm. Detailed Implementation
[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0024] Example
[0025] Combination Figure 1 As shown, a method for extracting multiple parameters of induced polarization using active source magnetization-time-domain electromagnetic detection includes:
[0026] 1) Separate the data for the flat-top phase and the shutdown time;
[0027] 2) The magnetic susceptibility parameter was extracted from the flat-top stage data using the active magnetic method;
[0028] 3) The extracted magnetic susceptibility is used as a constraint condition for time-domain electromagnetic method to extract resistivity and polarizability parameters;
[0029] 4) Obtain resistivity, polarizability and magnetic susceptibility parameters.
[0030] Furthermore, in step 2, combined with Figure 2 As shown, a parameter extraction model f = Gm is constructed, where m is the magnetic susceptibility, f is the response data, and G is the numerical simulation model operator; these are then substituted into the Occam algorithm. Where m ref Let C be the magnetic susceptibility of the reference model, C be the roughness matrix composed of second-order differential operators, and W be the magnetic susceptibility of the reference model. z Let d be a diagonal matrix, D be the diagonal weighted matrix of the reference model, and d be the diagonal matrix. obs These are the observed data, F(m) is the predicted response of the magnetic susceptibility model, and λ and μ are trade-off parameters; the Occam parameter extraction model update formula is...
[0031] The magnetic susceptibility χ is extracted through iterative looping.
[0032] Furthermore, in step 3, combined with Figure 3 As shown, the extracted magnetic susceptibility χ is used as a constraint condition to correct the response curve in the time-domain electromagnetic numerical simulation; a DQN parameter extraction model is constructed:
[0033] First, construct the objective function.
[0034]
[0035] Where (σ,η) represents the resistivity σ and polarizability η in the current state, (σ′,η′) represents the resistivity σ′ and polarizability η′ in the next state, (Δσ,Δη) represents the current action perturbation value, (Δσ′,Δη′) represents the next action perturbation value, and Q′ represents the updated state-action value. The action reward r is obtained in state (σ,η) and evaluated. γ is a discount factor; a larger γ emphasizes long-term rewards, while a smaller γ emphasizes short-term gains. θ is a neural network parameter. The target state-action value function y′=r+γmaxQ[(σ,η),(Δσ,Δη),θ] is constructed, where y′ represents the target Q value. The loss function L(θ)=E{{y′-Q[(σ,η),(Δσ,Δη),θ]} is constructed. 2}, where θ is the weight parameter trained in the neural network structure model.
[0036] Secondly, a Q-value network is constructed. DQN evaluates the current state-action value function through a target network and an estimation network. The target function obtains the target Q-value based on the neural network, and the target Q-value is used to estimate the Q-value at the next time step. The estimation network updates the network weights Δθ using stochastic gradient descent, where... It is the gradient operator, responsible for adjusting the parameters along the negative gradient direction. The gradient descent algorithm is...
[0037]
[0038] Finally, an experience replay pool is constructed. When the agent interacts with the environment, a sample database can be obtained. The sample database is stored in the established experience pool. A portion of the data is randomly selected from the experience pool for training samples, and then the training samples are fed into the neural network for training.
[0039] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for extracting multiple parameters of active source magnetization-time-domain electromagnetic induction polarization, characterized in that, Includes the following steps: 1) Separate the data for the flat-top phase and the shutdown time; 2) The magnetic susceptibility parameter was extracted from the flat-top stage data using the active magnetic method; 3) Using the extracted magnetic susceptibility as a known condition of the time-domain electromagnetic method, substitute it into the time-domain electromagnetic response stage after turn-off to extract resistivity and polarizability parameters. In step 3), the extracted magnetic susceptibility X is used as a constraint condition to correct the response curve in the time-domain electromagnetic numerical simulation; a DQN parameter extraction model is constructed: First, construct the objective function. ， Where (σ,η) represents the resistivity σ and polarizability η in the current state, (σ',η') represents the resistivity σ' and polarizability η' in the next state, (∆σ,∆η) represents the current action perturbation value, (∆σ',∆η') represents the next action perturbation value, Q' represents the updated state-action value, the action reward r is obtained in state (σ,η), and it is evaluated; γ is a discount factor, the larger γ is, the more emphasis is placed on long-term rewards, and the smaller γ is, the more emphasis is placed on short-term gains, and θ is the weight parameter trained in the neural network structure model; construct the target state-action value function. , where y' represents the target Q value; construct the loss function. , where θ is the weight parameter trained in the neural network structure model; Secondly, a Q-value network is constructed. DQN evaluates the current state-action value function through a target network and an estimation network. The objective function obtains the target Q value based on the neural network, and uses the target Q value to estimate the Q value at the next time step; The network is estimated to update its weights Δθ using stochastic gradient descent, where ▽ θ It is the gradient operator, responsible for adjusting parameters along the negative gradient direction; the gradient descent algorithm is... ； Finally, an experience replay pool is constructed. When the agent interacts with the environment, a sample database can be obtained. The sample database is stored in the established experience pool. A portion of the data is randomly extracted from the experience pool for training samples, and then the training samples are fed into the neural network for training. 4) Obtain resistivity, polarizability and magnetic susceptibility parameters.
2. The method for extracting multiple parameters of active source magnetization-time-domain electromagnetic induction polarization according to claim 1, characterized in that, In step 2, a parameter extraction model is constructed. Where m is the magnetic susceptibility, f is the response data, and G is the numerical simulation model operator; these are then substituted into the Occam algorithm. , where m ref Let C be the magnetic susceptibility of the reference model, C be the roughness matrix composed of second-order differential operators, and W be the magnetic susceptibility of the reference model. z Let d be a diagonal matrix, D be the diagonal weighted matrix of the reference model, and d be the diagonal matrix. obs These are the observed data, F(m) is the predicted response of the magnetic susceptibility model, and λ and μ are trade-off parameters; the Occam parameter extraction model update formula is... The magnetic susceptibility X is extracted through iterative loops.