Deep exploration method for altered rock type gold deposit based on time-frequency electromagnetic polarizability difference

Abstract

This invention discloses a method for deep gold exploration in altered rock-type deposits based on time-frequency electromagnetic polarizability differences, belonging to the field of mineral resource exploration technology. The invention first employs a broadband pseudo-random coded transmission system to generate multi-frequency superimposed excitation currents in the time domain and acquires time-domain voltage signals at the receiving end. After demodulation, voltage and current components at each frequency point are extracted, and the equivalent complex impedance of the underground medium is calculated to form a complex impedance sequence. Then, the Cole-Cole polarization parameter inversion with coding constraints is performed on this sequence to obtain the optimal polarizability. Based on this, the time-frequency polarization structure coupling difference, stability factor, and total spectral transition intensity are calculated separately, and these three are concatenated to construct a feature fusion vector, which is finally input into a classifier to automatically identify the mineralization category of the measuring point. This invention significantly improves the identification accuracy and anti-interference capability of deep, concealed mineralization anomalies and is suitable for gold exploration under complex geological conditions.

Description

Technical Field

[0001] This invention relates to the field of mineral resource exploration technology, specifically to a deep prospecting method for altered rock-type gold deposits based on the difference in time-frequency electromagnetic polarization. Background Technology

[0002] In the exploration of deep mineral resources, altered rock-type gold deposits, due to their deep burial and subtle differences in mineralization and surrounding rock properties, are among the most challenging targets in current geophysical exploration. While traditional direct current resistivity and frequency domain electromagnetic methods are widely used in shallow structural exploration, their resolution and detection depth are significantly limited when dealing with concealed ore bodies hundreds of meters or even deeper. In recent years, Spectral Induced Polarization (SIP) has been introduced into deep mineral exploration practices because it can identify mineralization anomalies by utilizing the dispersion characteristics of rock and ore polarizability at different frequencies. This method measures the complex resistivity spectrum over a wide frequency band and performs parameter inversion based on mathematical and physical models such as the Cole-Cole model, aiming to extract information related to the mineralized body from parameters such as polarizability, time constant, and frequency correlation coefficient.

[0003] However, traditional processing methods typically treat the complex impedance inversion results at each frequency point as independent parameter sets, focusing only on the absolute value anomalies of polarizability at a single frequency point or a few characteristic frequencies, while neglecting the overall structural characteristics of the polarization response of the subsurface medium in the frequency dimension. Specifically, current methods lack effective modeling of the intrinsic correlation between polarizability at different frequencies, making it difficult to distinguish between "polarization structure anomalies" caused by mineralization and "false spectral distortions" caused by near-surface electromagnetic coupling, measurement noise, or non-mineralized geological bodies (such as graphitized strata and clay layers). This interpretation mode based on isolated frequency point information leads to mineralization anomaly signals being easily submerged in strong interference backgrounds, resulting in severe ambiguity of the inverted Cole-Cole parameters. Ultimately, this results in low accuracy and high false alarm rates in identifying deep mineralization anomalies, failing to meet the reliability requirements of target area selection in actual exploration projects. Summary of the Invention

[0004] To address the aforementioned shortcomings in existing technologies, this invention provides a deep prospecting method for altered rock-type gold deposits based on time-frequency electromagnetic polarization differences. This method solves the problems in existing technologies where neglecting the polarization structure characteristics in the frequency dimension leads to difficulty in distinguishing mineralized anomalies from interference signals and low accuracy in identifying deep mineralization anomalies.

[0005] To achieve the aforementioned objectives, the technical solution adopted by this invention is: a method for deep prospecting of altered rock-type gold deposits based on time-frequency electromagnetic polarization differences, comprising: S1. A wideband pseudo-random coding transmission system is adopted to construct a multi-frequency superimposed excitation current in the time domain; S2. At the receiving end, the voltage signal in the time domain is acquired and demodulated. The voltage and current components at each frequency point are extracted, and the equivalent complex impedance of the underground medium at each frequency point is calculated to form a complex impedance sequence. S3. Encode and constrain the Cole-Cole polarization parameter inversion of the complex impedance sequence to obtain the optimal polarizability; S4. Obtain the time-frequency polarization structure coupling difference based on the frequency weight and optimal polarization rate of each frequency point; S5. Based on the optimal polarization at each frequency point, obtain the structural probability and calculate the stability factor; S6. Calculate the transition intensity based on the first-order difference of the polarization of adjacent frequency points, and obtain the total spectral transition intensity after frequency weighting and accumulation. S7. The time-frequency polarization structure coupling difference, stability factor and total spectral transition intensity are concatenated to obtain the feature fusion vector. The feature fusion vector is then input into the classifier to obtain the mineralization category discrimination result of the measurement point.

[0006] Furthermore, S1 specifically includes: generating an excitation current by superimposing multiple pseudo-random coded signals at different frequencies through a wideband pseudo-random coded transmission system, and determining the current amplitude corresponding to each frequency point based on the reference frequency, the reference current amplitude, and the spectral attenuation coefficient.

[0007] Furthermore, S2 includes the following sub-steps: S21. At the receiving end, the voltage signal in the time domain is acquired and demodulated by the pseudo-random coded signal synchronized with the transmitting end to obtain the voltage and current components corresponding to each frequency point. S22. The ratio of voltage component to current component at the same frequency point is taken as the equivalent complex impedance of the underground medium at the corresponding frequency point. S23. Construct a complex impedance sequence by taking the equivalent complex impedance of the underground medium at each frequency point.

[0008] Furthermore, S3 includes the following sub-steps: S31. Normalize the frequency of each frequency point to obtain the normalized angular frequency; S32. Calculate the energy integral of the pseudo-random coded signal at each frequency point over the total recording time, and add the energy integrals of each frequency point to obtain the total energy integral; S33. Use the ratio of the energy integral of each frequency point to the total energy integral as the coding weight; S34. Substitute the normalized angular frequency of each frequency point into the Cole-Cole model to obtain the theoretical complex impedance; S35. Using coding weights to weight the difference between the equivalent complex impedance and the theoretical complex impedance of the underground medium, and a parameter smoothing regularization term, a coding constraint inversion objective function is constructed. The minimum value of the coding constraint inversion objective function is solved to obtain the optimal polarizability.

[0009] Furthermore, S4 includes the following sub-steps: S41. Obtain the frequency weight for each frequency point; S42. Weight the frequency weights with the optimal polarizability of the corresponding frequency points and sum them up to obtain the weighted polarization energy; S43. Subtract the optimal polarization of two adjacent frequency points, divide by the natural logarithm of the frequency ratio of adjacent frequency points, and then accumulate by frequency weight to obtain the dispersion difference. S44. Multiply the dispersion difference by the weighted polarization energy to obtain the time-frequency polarization structure coupling difference.

[0010] Furthermore, the formula for calculating the frequency weight in S41 is as follows: , in, For the first Frequency weight of each frequency point For reference frequency, For the first The frequency of each frequency point For the first The frequency of each frequency point It is a logarithmic function. The number of frequency points, This refers to the frequency point number.

[0011] Furthermore, S5 includes the following sub-steps: S51. Add the optimal polarizations of each frequency point to obtain the total polarization; S52. The ratio of the optimal polarizability to the total polarizability at each frequency point is taken as the structural probability. S53. Calculate the polarization structure entropy for the structure probability at each frequency point; S54. Calculate the stability factor based on the polarization structure entropy.

[0012] Furthermore, the formula for calculating the polarization structure entropy in S53 is: , in, For polarization structure entropy, For the first The structural probability of each frequency point, where ln is a logarithmic function. For frequency points, The number of frequency points; The formula for calculating the stability factor in S54 is: , in, This is a stability factor.

[0013] Furthermore, S6 includes the following sub-steps: S61. Subtract the optimal polarizability of adjacent frequency points to obtain the first-order difference of polarizability; S62. Calculate the transition intensity based on the first-order difference of polarizability and the optimal polarizability at the current frequency. S63. The transition intensity is weighted and accumulated using the frequency weights of the frequency points to obtain the total spectral transition intensity.

[0014] Furthermore, the classifier in S7 consists of an input layer, a first fully connected layer, a second fully connected layer, and an output layer connected in sequence.

[0015] The beneficial effects of this invention are as follows: 1. This invention constructs features in three dimensions: time-frequency polarization structure coupling difference, stability factor, and total spectral transition intensity, and integrates them into a feature vector input to a classifier for discrimination. It makes full use of the inherent structural information of full-band polarizability data, effectively distinguishes false spectral distortions caused by near-surface electromagnetic coupling, measurement noise, and non-mineralized geological bodies, significantly improves the ability to distinguish between deep mineral-induced anomalies and interference signals, and reduces the false alarm rate.

[0016] 2. This invention introduces frequency-dimensional constraints by encoding and constraining the Cole-Cole polarization parameter inversion of complex impedance sequences, which effectively alleviates the problem of multiple solutions to parameters in traditional inversion methods. This makes the optimal polarization obtained by inversion more stable and reliable, providing a high-quality data foundation for subsequent feature extraction.

[0017] 3. This invention breaks away from the traditional approach of "isolating each frequency point" and quantitatively characterizes the dynamic characteristics of polarization change with frequency by calculating the transition intensity of polarization of adjacent frequency points and the total intensity of spectral transition. At the same time, it uses structural probability to calculate the stability factor and comprehensively explores the structural characteristics of the polarization response of underground media in the frequency dimension from multiple perspectives, thereby improving data utilization.

[0018] 4. This invention combines feature extraction with machine learning classifiers to form an end-to-end discrimination process, which can directly output mineralization category discrimination results at each measurement point, avoiding the subjectivity of relying on human experience for anomaly interpretation, and improving exploration efficiency and the objectivity of discrimination. Attached Figure Description

[0019] Figure 1 This is a flowchart of a deep prospecting method for altered rock-type gold deposits based on the difference in time-frequency electromagnetic polarization. Detailed Implementation

[0020] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0021] like Figure 1 As shown, a deep prospecting method for altered rock-type gold deposits based on time-frequency electromagnetic polarizability differences includes: S1. A wideband pseudo-random coding transmission system is adopted to construct a multi-frequency superimposed excitation current in the time domain; S2. At the receiving end, the voltage signal in the time domain is acquired and demodulated. The voltage and current components at each frequency point are extracted, and the equivalent complex impedance of the underground medium at each frequency point is calculated to form a complex impedance sequence. S3. Encode and constrain the Cole-Cole polarization parameter inversion of the complex impedance sequence to obtain the optimal polarizability; S4. Obtain the time-frequency polarization structure coupling difference based on the frequency weight and optimal polarization rate of each frequency point; S5. Based on the optimal polarization at each frequency point, obtain the structural probability and calculate the stability factor; S6. Calculate the transition intensity based on the first-order difference of the polarization of adjacent frequency points, and obtain the total spectral transition intensity after frequency weighting and accumulation. S7. The time-frequency polarization structure coupling difference, stability factor and total spectral transition intensity are concatenated to obtain the feature fusion vector. The feature fusion vector is then input into the classifier to obtain the mineralization category discrimination result of the measurement point.

[0022] In this embodiment, S1 specifically includes: generating an excitation current by superimposing multiple pseudo-random coded signals at different frequencies through a wideband pseudo-random coded transmission system, and determining the current amplitude corresponding to each frequency point based on the reference frequency, the reference current amplitude, and the spectral attenuation coefficient.

[0023] , in, For the first Constant excitation current, For the first The frequency point The pseudo-random coded signal corresponding to time 1. The time number, , For the first Current amplitude at each frequency point For the first The frequency of each frequency point The preset reference frequency, This is the spectral attenuation coefficient. For frequency points, Reference frequency The corresponding reference current amplitude.

[0024] Traditional spectral induced polarization methods often employ a frequency-domain point-by-point sweeping transmission approach, requiring the sequential output of single-frequency signals. This results in lengthy and time-consuming wideband measurements, and frequent switching of transmission states can introduce system instability. This invention, S1, utilizes pseudo-random coding technology to directly superimpose multiple signals from different frequency points in the time domain, forming a composite excitation current. This enables simultaneous excitation from low frequencies (such as...) with a single transmission. Hz) to high frequencies (such as The system can acquire complete spectral response information for all target frequency components across the entire frequency band (Hz) with a single observation. In actual field operations, this significantly shortens the observation time for a single measurement point, improves the efficiency of deep exploration, and avoids system drift errors caused by transmitter state switching during frequency-by-frequency transmission.

[0025] This invention uses formula Introducing spectral attenuation coefficient It can flexibly adjust the amplitude of the emitted current of different frequency components according to the electrical distribution characteristics and noise environment of the exploration target area. For example, deep mineralization information is usually carried in the low-frequency electromagnetic response, but low-frequency signals suffer severe energy attenuation during transmission and propagation, and are easily submerged by power frequency interference and ground noise. By setting... (If a negative value or a small positive value is taken), the low-frequency transmission energy can be actively increased and the high-frequency energy can be appropriately suppressed to achieve the optimal allocation of transmission power. The limited transmitter power can be concentrated on the key frequency band carrying useful geological information, effectively improving the excitation intensity and signal-to-noise ratio of weak signals in the deep environment.

[0026] In this embodiment, S2 includes the following sub-steps: S21. At the receiving end, the voltage signal in the time domain is acquired and demodulated by the pseudo-random coded signal synchronized with the transmitting end to obtain the voltage and current components corresponding to each frequency point. S22. The ratio of voltage component to current component at the same frequency point is taken as the equivalent complex impedance of the underground medium at the corresponding frequency point. S23. Construct a complex impedance sequence by taking the equivalent complex impedance of the underground medium at each frequency point.

[0027] In this embodiment, the demodulation formula in S21 is: , , in, For the first The equivalent voltage amplitude at each frequency point For the first Voltage signal at time, The total recording time of the pseudo-random coded signal. For the first The equivalent current amplitude at each frequency point For the first The excitation current is constantly applied.

[0028] The formula for the equivalent complex impedance of the underground medium in S22 is: , in, For the first The equivalent complex impedance of the underground medium at each frequency point.

[0029] The complex impedance sequence in S23 is: ,in, It is a complex impedance sequence. The equivalent complex impedance of the underground medium at the first frequency point, The equivalent complex impedance of the underground medium at the second frequency point, For the first Equivalent complex impedance of underground medium at each frequency point This represents the number of frequency points.

[0030] This invention performs correlation demodulation on time-domain voltage signals using pseudo-random coded signals synchronized at the receiving and transmitting ends. Utilizing the pulse characteristics of the autocorrelation function of the pseudo-random coded signal and the low cross-correlation between codes at different frequencies, it achieves coherent accumulation of useful signal energy during integration. Simultaneously, it effectively suppresses uncorrelated environmental electromagnetic noise, power frequency interference, and random ground fluctuations, significantly improving the signal-to-noise ratio of voltage and current components extracted at each frequency. This provides a reliable data foundation for extracting weak signals from deep underground environments in heavily interfered mining areas. Furthermore, by directly dividing the voltage and current components obtained from demodulation at the same frequency, the equivalent complex impedance of the underground medium is obtained. This ratio calculation naturally eliminates the influence of common-mode errors such as transmit current amplitude fluctuations, receiver system gain drift, and shallow non-uniform coupling, making the amplitude and phase information of the complex impedance at each frequency more accurate and reliable. Finally, the complex impedance across the entire frequency band is organized into a sequence, fully preserving the amplitude-frequency and phase-frequency response characteristics of the complex impedance at each frequency.

[0031] In this embodiment, S3 includes the following sub-steps: S31. Normalize the frequency of each frequency point to obtain the normalized angular frequency; S32. Calculate the energy integral of the pseudo-random coded signal at each frequency point over the total recording time, and add the energy integrals of each frequency point to obtain the total energy integral; S33. Use the ratio of the energy integral of each frequency point to the total energy integral as the coding weight; S34. Substitute the normalized angular frequency of each frequency point into the Cole-Cole model to obtain the theoretical complex impedance; S35. Using coding weights to weight the difference between the equivalent complex impedance and the theoretical complex impedance of the underground medium, and a parameter smoothing regularization term, a coding constraint inversion objective function is constructed. The minimum value of the coding constraint inversion objective function is solved to obtain the optimal polarizability.

[0032] In this embodiment, the formula for calculating the normalized angular frequency in S31 is: , in, For the first Normalized angular frequency at each frequency point This is the preset reference angular frequency.

[0033] The formula for the coding weight in S33 is: , in, For the first Encoding weights for each frequency point For the first The frequency point The pseudo-random coded signal corresponding to time 1. for The frequency point The pseudo-random coded signal corresponding to time 1. For differential operators, The total recording time of the pseudo-random coded signal. This represents the number of frequency points.

[0034] The expression for the Cole-Cole model in S34 is: , in, For theoretical complex impedance, Resistivity at zero frequency limit Polarizability Let be the relaxation time constant. The frequency correlation coefficient, The imaginary unit, For the first Normalized angular frequency at each frequency point , This is the Cole-Cole polarization parameter vector.

[0035] The expression for the coding constraint inversion objective function in S35 is: , in, For the first Equivalent complex impedance of underground medium at each frequency point For theoretical complex impedance, For the first Encoding weights for each frequency point These are the regularization weight coefficients. For parameter smoothing regularization, || is the absolute value operation. To Take norm, For frequency points, The number of frequency points, To minimize the value, the parameter smoothing regularization term is specifically the second-order difference norm of the model parameter vectors at adjacent frequency points.

[0036] This invention introduces a Cole-Cole polarization parameter inversion strategy with coding constraints. It constructs coding weights using the energy integral of pseudo-random coded signals at each frequency point. In the inversion objective function, it assigns greater weight to the complex impedance fitting residuals of frequencies with high signal-to-noise ratios and concentrated energy, allowing these frequencies, which carry more effective information, to play a dominant role in the parameter optimization process. At the same time, by introducing a parameter smoothing regularization term to constrain drastic changes between adjacent parameters, it effectively suppresses the polarization multiple solutions problem caused by uneven frequency data quality and model parameter overfitting in traditional inversion. This ensures that the optimal polarization obtained by inversion is faithful to the measured data of high-quality frequencies, significantly improving the accuracy and stability of the polarization parameter.

[0037] In this embodiment, S4 includes the following sub-steps: S41. Obtain the frequency weight for each frequency point; S42. Weight the frequency weights with the optimal polarizability of the corresponding frequency points and sum them up to obtain the weighted polarization energy; S43. Subtract the optimal polarization of two adjacent frequency points, divide by the natural logarithm of the frequency ratio of adjacent frequency points, and then accumulate by frequency weight to obtain the dispersion difference. S44. Multiply the dispersion difference by the weighted polarization energy to obtain the time-frequency polarization structure coupling difference.

[0038] In this embodiment, the formula for calculating the frequency weight in S41 is: , in, For the first Frequency weight of each frequency point For reference frequency, For the first The frequency of each frequency point For the first The frequency of each frequency point It is a logarithmic function. The number of frequency points, This refers to the frequency point number.

[0039] The formula for calculating the weighted polarization energy in S42 is: , in, For weighted polarization energy, For the first The optimal polarization rate at each frequency point.

[0040] The formula for calculating the dispersion difference in S43 is: , in, For dispersion differences, For the first Optimal polarization at each frequency point It is a logarithmic function.

[0041] This invention first uses logarithmic frequency weighting to weight and accumulate the polarization of each frequency point to obtain weighted polarization energy. This weighting form allows the low-frequency band to receive higher weights due to the stronger response of the geological body, effectively highlighting deep mineralization information. At the same time, the local change gradient of the spectral curve is measured by the ratio of the difference in polarization of adjacent frequency points to the logarithm of the frequency interval, and the dispersion difference is obtained by weighted accumulation with frequency weights, quantitatively characterizing the overall dispersion intensity of polarization as frequency changes. Finally, the two are multiplied to form a coupling difference, so that only spectral structures with both high polarization energy and strong dispersion characteristics can obtain a high coupling difference value, while single-dimensional anomalies (such as only high polarization but no dispersion, or only strong dispersion but low polarization) are effectively suppressed, thereby significantly enhancing the distinction between mineral-induced polarization structure anomalies and non-mineral-induced interference signals.

[0042] In this embodiment, S5 includes the following sub-steps: S51. Add the optimal polarizations of each frequency point to obtain the total polarization; S52. The ratio of the optimal polarizability to the total polarizability at each frequency point is taken as the structural probability. S53. Calculate the polarization structure entropy for the structure probability at each frequency point; S54. Calculate the stability factor based on the polarization structure entropy.

[0043] In this embodiment, the formula for calculating the polarization structure entropy in S53 is: , in, For polarization structure entropy, For the first The structural probability of each frequency point, where ln is a logarithmic function. For frequency points, The number of frequency points; The formula for calculating the stability factor in S54 is: , in, This is a stability factor.

[0044] In this invention, when the polarizability spectrum caused by mineralization exhibits a wide-bandwidth, continuous, and stable response characteristic, the polarizability at each frequency point is relatively uniformly distributed, the structural probability tends to be consistent, the polarization structure entropy is relatively high, and the corresponding stability factor remains at a reasonably low level, reflecting that the spectral structure of the mineralized polarization response has a stable physical origin. Conversely, when the measurement is interfered with by noise or encounters non-mineralized geological bodies, the polarizability spectrum often shows drastic jumps or random fluctuations at local frequency points, the structural probability distribution is extremely uneven, the polarization structure entropy is significantly reduced, and the stability factor increases accordingly, thereby effectively identifying abnormal spectral morphology.

[0045] In this embodiment, S6 includes the following sub-steps: S61. Subtract the optimal polarizability of adjacent frequency points to obtain the first-order difference of polarizability; S62. Calculate the transition intensity based on the first-order difference of polarizability and the optimal polarizability at the current frequency. S63. The transition intensity is weighted and accumulated using the frequency weights of the frequency points to obtain the total spectral transition intensity.

[0046] The formula for calculating the first-order difference of polarizability in S61 is: , in, For the first First-order difference of polarization at each frequency point For the first Optimal polarization at each frequency point For the first Optimal polarization at each frequency point; The formula for calculating the transition strength in S62 is: , in, For the first The transition intensity at each frequency point, where || represents the absolute value. It is a very small constant; The formula for calculating the total spectral transition intensity in S63 is: , in, The total spectral transition intensity, For the first Frequency weights of each frequency point; The polarizability spectrum caused by mineralization usually exhibits a regular dispersion pattern with smooth frequency variation. The polarizability changes gradually between adjacent frequency points, the relative transition intensity is small, and the total spectral transition intensity is at a low level. However, near-surface electromagnetic coupling, measurement spike noise, or non-mineralized highly polarizable bodies (such as graphitized strata and clay layers) often cause discontinuous and drastic jumps or oscillations in the spectral curve in local frequency bands. The relative rate of change between adjacent frequency points increases significantly, and the total spectral transition intensity increases sharply, thus effectively capturing local anomalous abrupt changes in the spectral structure. At the same time, the introduction of frequency weights reasonably amplifies the transition contribution of the low-frequency band (where the polarization response of geological bodies is more stable and mineralization information is richer), while the noise-induced jumps of the high-frequency band (which is easily affected by electromagnetic coupling) are relatively suppressed. This makes the total spectral transition intensity sensitive to local anomalous abrupt changes in the spectral curve, while also focusing on the transition changes in the key low-frequency bands.

[0047] In this embodiment, the classifier in S7 includes, in sequence, an input layer, a first fully connected layer, a second fully connected layer, and an output layer.

[0048] When inputting into the classifier, the time-frequency polarization structure coupling difference and the total spectral transition intensity in the feature fusion vector are normalized, with the stability factor ranging from 0 to 1.

[0049] This invention constructs features in three dimensions: time-frequency polarization structure coupling difference, stability factor, and total spectral transition intensity, and integrates them into a feature vector input to a classifier for discrimination. It makes full use of the inherent structural information of full-band polarizability data, effectively distinguishes false spectral distortions caused by near-surface electromagnetic coupling, measurement noise, and non-mineralized geological bodies, significantly improves the ability to distinguish deep mineral-induced anomalies from interference signals, and reduces the false alarm rate.

[0050] This invention introduces frequency-dimensional constraints by encoding and constraining the Cole-Cole polarization parameter inversion of complex impedance sequences, effectively alleviating the problem of multiple solutions to parameters in traditional inversion methods. This makes the optimal polarization obtained by inversion more stable and reliable, providing a high-quality data foundation for subsequent feature extraction.

[0051] This invention breaks away from the traditional approach of "isolating each frequency point" and quantitatively characterizes the dynamic characteristics of polarization as a function of frequency by calculating the transition intensity of polarization between adjacent frequency points and the total intensity of spectral transitions. At the same time, it uses structural probability to calculate the stability factor and comprehensively explores the structural characteristics of the polarization response of underground media in the frequency dimension from multiple perspectives, thereby improving data utilization.

[0052] This invention combines feature extraction with machine learning classifiers to form an end-to-end discrimination process, which can directly output mineralization category discrimination results at each measurement point, avoiding the subjectivity of relying on human experience for anomaly interpretation, and improving exploration efficiency and the objectivity of discrimination.

[0053] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for deep prospecting of altered rock-type gold deposits based on time-frequency electromagnetic polarizability differences, characterized in that, include: S1. A wideband pseudo-random coding transmission system is adopted to construct a multi-frequency superimposed excitation current in the time domain; S2. At the receiving end, the voltage signal in the time domain is acquired and demodulated. The voltage and current components at each frequency point are extracted, and the equivalent complex impedance of the underground medium at each frequency point is calculated to form a complex impedance sequence. S3. Encode and constrain the Cole-Cole polarization parameter inversion of the complex impedance sequence to obtain the optimal polarizability; S4. Obtain the time-frequency polarization structure coupling difference based on the frequency weight and optimal polarization rate of each frequency point; S5. Based on the optimal polarization at each frequency point, obtain the structural probability and calculate the stability factor; S6. Calculate the transition intensity based on the first-order difference of the polarization of adjacent frequency points, and obtain the total spectral transition intensity after frequency weighting and accumulation. S7. Combine the time-frequency polarization structure coupling difference, stability factor, and total spectral transition intensity. The feature fusion vector is obtained and input into the classifier to obtain the mineralization category discrimination result of the measurement point.

2. The method for deep exploration of altered rock-type gold deposits based on time-frequency electromagnetic polarization differences according to claim 1, characterized in that, S1 specifically includes: generating an excitation current by superimposing multiple pseudo-random coded signals at different frequencies through a wideband pseudo-random coded transmission system, and determining the current amplitude corresponding to each frequency point based on the reference frequency, the reference current amplitude, and the spectral attenuation coefficient.

3. The method for deep exploration of altered rock-type gold deposits based on time-frequency electromagnetic polarization differences according to claim 1, characterized in that, S2 includes the following steps: S21. At the receiving end, the voltage signal in the time domain is acquired and demodulated by the pseudo-random coded signal synchronized with the transmitting end to obtain the voltage and current components corresponding to each frequency point. S22. The ratio of voltage component to current component at the same frequency point is taken as the equivalent complex impedance of the underground medium at the corresponding frequency point. S23. Construct a complex impedance sequence by taking the equivalent complex impedance of the underground medium at each frequency point.

4. The method for deep exploration of altered rock-type gold deposits based on time-frequency electromagnetic polarization differences according to claim 1, characterized in that, S3 includes the following steps: S31. Normalize the frequency of each frequency point to obtain the normalized angular frequency; S32. Calculate the energy integral of the pseudo-random coded signal at each frequency point over the total recording time, and add the energy integrals of each frequency point to obtain the total energy integral; S33. Use the ratio of the energy integral of each frequency point to the total energy integral as the coding weight; S34. Substitute the normalized angular frequency of each frequency point into the Cole-Cole model to obtain the theoretical complex impedance; S35. Using coding weights to weight the difference between the equivalent complex impedance and the theoretical complex impedance of the underground medium, and a parameter smoothing regularization term, a coding constraint inversion objective function is constructed. The minimum value of the coding constraint inversion objective function is solved to obtain the optimal polarizability.

5. The method for deep exploration of altered rock-type gold deposits based on time-frequency electromagnetic polarization differences according to claim 1, characterized in that, S4 includes the following steps: S41. Obtain the frequency weight for each frequency point; S42. Weight the frequency weights with the optimal polarizability of the corresponding frequency points and sum them up to obtain the weighted polarization energy; S43. Subtract the optimal polarization of two adjacent frequency points, divide by the natural logarithm of the frequency ratio of adjacent frequency points, and then accumulate by frequency weight to obtain the dispersion difference. S44. Multiply the dispersion difference by the weighted polarization energy to obtain the time-frequency polarization structure coupling difference.

6. The method for deep exploration of altered rock-type gold deposits based on time-frequency electromagnetic polarization differences according to claim 5, characterized in that, The formula for calculating the frequency weight in S41 is as follows: ， in, For the first Frequency weight of each frequency point For reference frequency, For the first The frequency of each frequency point For the first The frequency of each frequency point It is a logarithmic function. The number of frequency points, This refers to the frequency point number.

7. The method for deep exploration of altered rock-type gold deposits based on time-frequency electromagnetic polarization differences according to claim 1, characterized in that, S5 includes the following steps: S51. Add the optimal polarizations of each frequency point to obtain the total polarization; S52. The ratio of the optimal polarizability to the total polarizability at each frequency point is taken as the structural probability. S53. Calculate the polarization structure entropy for the structure probability at each frequency point; S54. Calculate the stability factor based on the polarization structure entropy.

8. The method for deep exploration of altered rock-type gold deposits based on time-frequency electromagnetic polarization differences according to claim 7, characterized in that, The formula for calculating the polarization structure entropy in S53 is as follows: ， in, For polarization structure entropy, For the first The structural probability of each frequency point, where ln is a logarithmic function. The frequency point number, The number of frequency points; The formula for calculating the stability factor in S54 is: ， in, This is a stability factor.

9. The method for deep exploration of altered rock-type gold deposits based on time-frequency electromagnetic polarization differences according to claim 1, characterized in that, S6 includes the following sub-steps: S61. Subtract the optimal polarizability of adjacent frequency points to obtain the first-order difference of polarizability; S62. Calculate the transition intensity based on the first-order difference of polarizability and the optimal polarizability at the current frequency. S63. The transition intensity is weighted and accumulated using the frequency weights of the frequency points to obtain the total spectral transition intensity.

10. The method for deep prospecting of altered rock-type gold deposits based on time-frequency electromagnetic polarization differences according to claim 1, characterized in that, The classifier in S7 comprises, in sequence: an input layer, a first fully connected layer, a second fully connected layer, and an output layer.

Images