A data prediction method for achieving high-resolution subsurface layering in deep exploration

Abstract

This invention provides a high-resolution data prediction method for subsurface stratification in deep earth exploration. It utilizes a signal transmitter and receiver to conduct multi-mode geological scanning of the subsurface, acquiring raw subsurface reflection signal data. The acquired raw subsurface reflection signal data undergoes preprocessing, time-frequency transformation, and feature extraction to obtain time-domain and frequency-domain feature parameters. These extracted time-domain and frequency-domain feature parameters are input into a pre-constructed coupled prediction model for sequential calculation, outputting prediction results for the depth range and material type of subsurface stratification. The prediction results are compared and verified with data in a known geological database to generate a final report on the subsurface stratification structure and material composition. This invention overcomes the shortcomings of existing deep earth exploration technologies, such as incompatibility between depth and resolution, low accuracy in material identification, and lack of confidence assessment, improving the accuracy and reliability of subsurface stratification and material prediction, and providing precise data support for deep earth resource exploration and geological surveys.

Description

Technical Field

[0001] This invention relates to the field of geological exploration, and more particularly to a method for high-resolution data prediction of subsurface stratification in deep earth exploration. Background Technology

[0002] Geological exploration technology is a core means for humans to understand underground space and develop deep-earth resources. Early geological exploration relied on manual drilling and simple geophysical methods (such as gravity exploration and magnetic exploration), which could only achieve a preliminary judgment of macroscopic geological structures. The exploration depth was mostly limited to the hundreds of meters underground, and it could not obtain fine information on stratigraphic layering. In the mid-20th century, seismic exploration gradually became the mainstream technology for deep-earth exploration. By analyzing the differences in the propagation speed of seismic waves in different strata, it can achieve structural exploration thousands of meters underground and has been widely used in the field of oil and gas resource exploration. However, this technology relies on artificial seismic sources, which causes great environmental damage, and has extremely low resolution, making it difficult to identify fine structures such as thin interbedded layers and small-scale karst caves.

[0003] With the development of electronic technology, high-frequency electromagnetic wave detection methods (such as ground-penetrating radar) emerged in the late 20th century. By emitting high-frequency electromagnetic pulses, it utilizes the difference in dielectric constant of different strata to form reflected signals, achieving high-resolution detection at the centimeter level. It is suitable for shallow engineering geological exploration (such as roadbed and foundation detection). However, due to the strong attenuation characteristics of high-frequency signals, its detection depth is usually less than 10 meters, which cannot meet the needs of deep-earth exploration.

[0004] In recent years, the industry's demand for deep exploration and high resolution has become increasingly urgent. However, existing technologies have always faced a core contradiction: the incompatibility between depth and resolution. While seismic exploration methods can meet the needs of deep-earth exploration, insufficient resolution leads to the omission of key fine structures. High-frequency electromagnetic wave exploration methods, although having high resolution, have limited exploration depth and cannot cover deep-earth scenarios. In addition, existing technologies mostly rely on single signal features (such as seismic wave velocity and electromagnetic wave amplitude) for stratigraphic inversion, lacking the fusion analysis of multi-dimensional signal features, resulting in low accuracy in identifying stratigraphic material types (such as the inability to distinguish between sandstone and cement-bearing rock). At the same time, most methods can only output stratigraphic depth information and lack confidence assessment of the prediction results, making it difficult to support the reliability requirements of engineering decisions. These problems collectively restrict the application effectiveness of deep-earth exploration technology in resource development, engineering safety, and other fields.

[0005] Therefore, there is an urgent need in this field for a data prediction method that can achieve high resolution of subsurface stratification in deep earth exploration in order to solve the above problems. Summary of the Invention

[0006] The purpose of this application is to overcome the shortcomings of existing deep-earth exploration technologies, such as incompatibility between depth and resolution, low accuracy of material identification, and lack of confidence assessment. It provides a data prediction method for achieving high-resolution subsurface stratification in deep-earth exploration. This method acquires high-quality reflection signals through multi-mode geological scanning, extracts multi-dimensional feature parameters by combining time-frequency transformation, constructs a coupled mathematical prediction model to invert the dielectric constant distribution and material type, and finally verifies the stratification results with confidence by comparing with known geological databases. This achieves the synergy of large exploration depth and high resolution in deep-earth scenarios, improving the accuracy and reliability of subsurface stratification and material prediction, and providing precise data support for deep-earth resource exploration and engineering geological investigation.

[0007] This invention provides a method for high-resolution subsurface stratification data prediction in deep earth exploration, comprising the following steps:

[0008] Step 1: Use a signal transmitter and receiver to conduct multi-mode geological scanning of the underground to collect raw underground reflection signal data;

[0009] Step 2 involves preprocessing, time-frequency transformation, and feature extraction of the collected raw underground reflection signal data to obtain time-domain and frequency-domain feature parameters.

[0010] Step 3: Input the extracted time-domain and frequency-domain feature parameters into the pre-built coupled prediction model for sequential calculation, and output the prediction results of the depth range and material type of the subsurface layer;

[0011] Step four involves comparing and verifying the prediction results obtained in step three with data from known geological databases to generate the final report on underground layered structure and material composition.

[0012] Compared with the prior art, the beneficial effects of this application are as follows:

[0013] 1. This application employs a composite pulse beam with adjustable group velocity (pulse radio waves, microwave beams, and multi-mode scanning (WARR-Scan, CMP-Scan, etc.)) to achieve deep underground layered detection using low-frequency signals and ensure resolution by fusing high-frequency signals with multiple features, thus solving the core contradiction of low resolution in seismic detection methods and shallow depth in high-frequency electromagnetic wave detection methods.

[0014] 2. This application constructs a coupled prediction model of dielectric constant inversion (first sub-model) - material probability matching (second sub-model). The dielectric constant is inverted through multi-dimensional features such as energy spectrum, group delay, and instantaneous frequency. The material type is matched based on the normal distribution probability model, which can effectively tolerate the natural fluctuation of dielectric constant of similar materials and improve the material identification accuracy of complex strata (such as mixed lithology and cement-bearing rocks).

[0015] 3. This application introduces a confidence index calculation formula, which integrates the two-dimensional evaluation of material probability matching degree and depth deviation rationality. By comparing with the reference depth and material parameters of known geological databases, it outputs the confidence value of each layer, which solves the problem that the existing technology only outputs deterministic results and lacks reliability assessment, and provides a quantitative basis for engineering decisions (such as drilling target selection and tunnel risk prediction).

[0016] 4. The multi-mode scanning of this application can be flexibly selected according to the exploration needs (such as shallow engineering exploration and deep resource exploration). The model parameters can be calibrated and adapted to the stratigraphic characteristics through known samples. It does not rely on fixed seismic sources or complex equipment. The scope of application covers multiple scenarios such as oil and gas exploration, mineral development, and engineering geology, which reduces the equipment cost and operational complexity of deep earth exploration.

[0017] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0019] Figure 1 This is a flowchart illustrating a method for high-resolution subsurface stratification data prediction in deep earth exploration, as provided in an embodiment of the present invention. Detailed Implementation

[0020] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] Example 1:

[0022] This invention provides a method for high-resolution subsurface stratification data prediction in deep earth exploration. Please refer to [link to relevant documentation]. Figure 1 This includes the following steps:

[0023] Step 1: Use a signal transmitter and receiver to conduct multi-mode geological scanning of the underground to collect raw underground reflection signal data;

[0024] In step one, the multi-mode geological scanning includes at least one of the following scanning methods: wide-angle reflection and refraction scanning (WARR-Scan), profile scanning (Profile-Scan), static staring scanning (Stear-Scan), and common midpoint scanning (CMP-Scan); the signal transmitter transmits pulsed radio waves and microwave beam signals to the underground through a transmitting antenna, and the group velocity of the pulsed radio waves and microwave beam signals can be adjusted by the frequency adjustment module built into the transmitter; the receiver receives the reflected signals returned after being acted upon by underground materials through a receiving antenna.

[0025] Specifically, the Wide Angle Relection Refraction (WARR-Scan) method involves: separately setting the transmitting and receiving antennas at a fixed starting point; exciting the transmitting antenna to move along a predetermined path on the ground from the starting point, away from the receiving antenna fixed at the starting point; during the movement of the transmitting antenna, as the separation distance between the transmitting and receiving antennas increases, periodically recording the position of the transmitting antenna and the distance between the transmitting and receiving antennas; and receiving the continuously reflected signals returned after the pulse beam signals emitted by the transmitting antenna at each of the stated position points interact with the underground material.

[0026] The Profile Scan method is as follows: the transmitting antenna and the receiving antenna are set separately at a fixed starting point, and the distance between the transmitting antenna and the receiving antenna is fixed; the transmitting antenna and the receiving antenna are controlled to move simultaneously along the measurement line on the ground surface. During the movement, the pulse beam signal generated by the transmitting antenna is transmitted vertically or at a non-vertical angle to the ground. Each transmitted pulse beam signal is reflected from the feature surface underground and generates continuous transmission signal tracking on the receiving antenna.

[0027] The static staring scan method involves setting up transmitting and receiving antennas at multiple fixed points. At each fixed point, the transmitting antenna transmits pulse beam signals into the ground while the receiving antenna collects the reflected signals returned after the pulse beam signals interact with the underground material within a certain time period, in order to collect underground depth and layer data.

[0028] The midpoint CMP scanning method is as follows: the transmitting antenna and the receiving antenna are moved in opposite directions from a common point at the same or the same speed. At each fixed point during the movement of the transmitting antenna and the receiving antenna, the reflected signal corresponding to the pulse beam signal emitted by the transmitting antenna is collected to obtain underground depth and layer data; wherein, the reflection point of the reflected signal corresponding to each fixed point is located on a vertical line.

[0029] Step 2 involves preprocessing, time-frequency transformation, and feature extraction of the collected raw underground reflection signal data to obtain time-domain and frequency-domain feature parameters.

[0030] The preprocessing includes denoising, normalization, and time-frequency transformation. Denoising employs wavelet transform, and normalization uses min-max normalization. Time-frequency transformation uses short-time Fourier transform, employing a sliding window function to transform the original time-domain reflection signal. The signal is divided into several short-time segments, and a Fourier transform is performed on each segment to obtain a complex-valued signal in the frequency-time domain. and phase spectrum Then, based on the propagation law of electromagnetic waves, the time dimension t is mapped to the underground depth dimension z to obtain the frequency-depth domain complex signal. Further extraction of energy spectrum Group delay Instantaneous frequency , i.e., time domain and frequency domain characteristic parameters;

[0031] The mapping logic for the time dimension t to the subsurface depth dimension z is as follows: obtain the local theoretical average dielectric constant of the formation in the detection area. As an initial approximation (which is known and can be obtained from a local database, so it will not be described in detail further), it is based on the initial value of the propagation speed of electromagnetic waves in the underground medium. ( (The speed of light in a vacuum), combined with the time relationship of two-way signal propagation. To obtain the initial depth mapping relationship Thus Converted into an initial frequency-depth domain complex signal Its energy spectrum is ,Right now , characterizing the energy intensity of the reflected signal at a depth z and a frequency f;

[0032] Group delay Instantaneous frequency The method for determining it is as follows:

[0033]

[0034] in, This represents the phase spectrum obtained after Fourier transforming the original underground reflected signal. The time-domain expression representing the original underground reflected signal. Represents the time-domain reflected signal Take the argument angle, Indicates signal frequency. Indicates time.

[0035] Step 3: Input the extracted time-domain and frequency-domain feature parameters into the pre-built coupled prediction model for sequential calculation, and output the prediction results of the depth range and material type of the subsurface layer;

[0036] The coupled prediction model consists of a first prediction sub-model and a second prediction sub-model.

[0037] The first prediction sub-model is used to calculate the dielectric constant distribution at different depths underground based on the time-domain and frequency-domain feature parameters extracted in step two. The first prediction sub-model is as follows:

[0038]

[0039] in, denoted by , representing the dielectric constant distribution, indicating the relative dielectric constant at depth z below ground; A is the model gain coefficient, obtained through calibration using known geological samples; B is the attenuation coefficient, determined based on the stratigraphic characteristics of the detection area. The frequency decay exponent was obtained by fitting experimental data.

[0040] Specifically, the process of obtaining the model gain coefficient A is as follows: Select a known geological sample (such as a stratigraphic core obtained through drilling) that is similar to the geological conditions of the detection area, and measure its actual dielectric constant at different frequencies and depths; at the same time, collect the reflection signal of the sample area, extract the energy spectrum, group delay and instantaneous frequency according to step two, substitute them into the formula of the first prediction sub-model, and adjust the value of A by the least squares method (or gradient descent method) to minimize the sum of squared errors between the model calculation value and the actual measurement value (such as the error being less than 5%). At this time, A is the calibrated value.

[0041] The specific process for obtaining the attenuation coefficient B is as follows: Based on the stratigraphic characteristics of the detection area (such as stratigraphic homogeneity, porosity, water content, etc.), an initial range is preset. The specific steps are as follows: If the detection area is a relatively homogeneous dense rock layer (such as limestone), the initial value of B is taken as a small value (such as 0.01-0.03 ns / Hz); if it is a loose sedimentary layer (such as sandy soil layer), the initial value of B is taken as a large value (such as 0.05-0.1 ns / Hz); combined with the statistical range of B values ​​for similar strata in the known geological database, the specific value of B is finally determined by 1-2 field test measurements (comparing the model output with local borehole data).

[0042] Frequency attenuation index The specific acquisition process is as follows: In a laboratory environment, typical geological materials (such as sandstone and shale) that may exist in the detection area are simulated, and the attenuation law of the energy spectrum of the reflected signal at different frequencies is measured; the experimental data are then processed according to... The relationship is fitted (using log-linear regression) to obtain the fitted value of α; the fitted values ​​of α for multiple materials are weighted and averaged (the weights are determined according to the probability of stratigraphic distribution in the exploration area) to finally obtain the α value of the model;

[0043] The core principle of the first predictive sub-model is based on the interaction between electromagnetic waves and the underground medium. It uses the time-frequency characteristics of the signal to invert the dielectric constant distribution of the medium, where the dielectric constant of the underground medium... It is a core electromagnetic parameter reflecting material properties, and is directly related to the propagation speed and reflection characteristics of electromagnetic waves. The larger the value, the slower the electromagnetic wave propagation speed, and the faster the reflected signal energy attenuates; energy spectrum The signal energy intensity at different frequencies and depths is directly related to the absorption and attenuation characteristics of the medium for electromagnetic waves; instantaneous frequency Characterizing the change of signal frequency over time, it is related to the abrupt change characteristics of reflected signals caused by the medium delamination interface; group delay. The propagation time of the signal energy envelope is described by its relation to the propagation speed of electromagnetic waves in the medium (and thus to...). Directly related; through integration, the characteristics of the frequency domain ( The dielectric constant of the depth domain Correlation, where the index term Used to quantize the effect of group delay on signal energy attenuation; The term is used to describe the natural attenuation law of high-frequency signals propagating underground; the gain coefficient A is used to calibrate the overall output scale of the model to ensure that it matches the actual dielectric constant range; the time domain (instantaneous frequency) and frequency domain (energy spectrum, group delay) features are fused through the coupling integral formula, which breaks through the limitations of traditional single feature inversion and realizes high-resolution calculation of dielectric constant.

[0044] The second predictor sub-model is used to predict the dielectric constant distribution output by the first predictor sub-model. The second prediction sub-model is used to predict material types at different depths underground.

[0045]

[0046] in, This indicates that the relative permittivity at the measured depth z is... In this case, the medium at that depth belongs to the i-th type of material. The probability of is in the range of [0,1]. This represents the predefined i-th type of underground material; The mean dielectric constant of the i-th type of material is obtained by statistical analysis of dielectric constant data of this type of material in a known geological database. The standard deviation of the dielectric constant of the i-th type of material is calculated from the dielectric constant data of this type of material in a known geological database.

[0047] Specifically, Extracting dielectric constant sample data for material type i (such as sandstone, shale, etc.) from a known geological database involves the following steps: collecting dielectric constant measurements of this type of material under different geological conditions (such as temperature, pressure, and water content), with a sample size of no less than 30 sets (to ensure statistical significance); processing outliers in the sample data (e.g., removing extreme values ​​that deviate from the mean by more than three standard deviations); and calculating the arithmetic mean of the processed samples, i.e. ,in( The dielectric constant sample values ​​of the i-th type of material, where n is the number of valid samples;

[0048] Based on the above average The sample data is calculated through the following steps: calculate the squared deviation of each sample value from the mean; calculate the average of the squared deviations (i.e., the variance); take the square root of the deviation to obtain the standard deviation, which is used to characterize the dispersion of the dielectric constant of this type of material (the calculation method is based on mathematical principles and will not be elaborated in detail).

[0049] The second prediction sub-model is based on the core idea of ​​matching probability statistics with the electromagnetic properties of materials. It achieves prediction through the mapping relationship between dielectric constant and material type. The dielectric constants of different underground materials (such as rocks, soils, and ores) have significant statistical differences, and the dielectric constants of similar materials statistically follow a normal distribution (affected by minor differences in composition, structure, etc.). Therefore, this embodiment uses a normal distribution probability density function to describe the correlation between dielectric constant and material type, that is, for the dielectric constant at a certain depth z... Calculate the probability that it belongs to the i-th type of material Mᵢ. ;when Approximate to the average value of this type of material When, probability value Maximum; when deviating At that time, the probability decreases exponentially with the square of the deviation (from (Controlling the decay rate); By comparing the probability values ​​of different material types, the material type with the highest probability is used as the prediction result at that depth, realizing the mapping between dielectric constant and material type; By introducing probability distribution instead of deterministic threshold for material matching, the natural fluctuation of dielectric constant of similar materials is effectively tolerated, and the prediction robustness of complex strata (such as mixed lithological areas) is improved.

[0050] The process of outputting the predicted depth range and material type of underground strata is as follows: based on the material type probability distribution output by the second prediction sub-model. The location where the predicted material type changes at adjacent depths is taken as the stratification interface. The depth range between the interfaces is the depth range of a subsurface stratification, and each stratification corresponds to a unique predicted material type.

[0051] Step four involves comparing and verifying the prediction results obtained in step three with data from known geological databases to generate the final report on underground layered structure and material composition.

[0052] The comparison and verification process includes:

[0053] The subsurface stratification depth range and predicted material type of each stratum determined in step three are matched with standard stratigraphic data of the same or similar areas in the known geological database, and the confidence index of the prediction result of each stratum is calculated.

[0054] The final generated report on underground layered structure and material composition includes a visualized underground layered profile, and labels the predicted material type, depth range and corresponding confidence index for each layer;

[0055] The confidence index is calculated using the following formula:

[0056]

[0057] in, This represents the confidence level of the prediction result at a subsurface depth z, with a value range of [0,1]. This represents the maximum probability among all material types at depth z. This indicates the reference depth of the corresponding stratum in a known geological database; This represents the confidence attenuation coefficient, which is determined based on the geological stability of the detection area.

[0058] The overall confidence level for each subsurface stratum is the sum of the values ​​at all z locations within the depth range of that stratum. The average value.

[0059] Specifically, the reference depth is the standard stratigraphic interface depth in a known geological database that is similar to the stratigraphic characteristics of the exploration area. The method for obtaining this depth is as follows: Data from regions in the database that are closest to the geological structure (e.g., stratigraphic chronology, sedimentary environment) of the exploration area are selected, and their verified stratigraphic interface depths (e.g., depths calibrated through borehole measurements or seismic exploration) are extracted. If multiple similar regional data exist, the arithmetic mean of the corresponding stratigraphic interface depths for each region is taken as the reference depth. For cases where there is no complete match in the database, the reference depth is estimated by geological analogy (such as stratigraphic correlation of adjacent areas), and the range of uncertainty is marked.

[0060] The confidence decay coefficient is used to quantify the impact of the deviation between the predicted depth and the reference depth on the confidence level. Its specific determination depends on the formation stability; that is, for tectonically stable strata (such as sedimentary rocks undisturbed by faults), the lateral continuity of the stratigraphic interfaces is good, and the tolerance for deviation is high. Take the larger value; for tectonically active strata (such as fault-developed areas and volcanic rock distribution areas), stratigraphic interfaces are prone to abrupt changes, and the tolerance for deviation is low. Take the smaller value; finally, calibrate using a validation sample (a dataset with known deviations between predicted and actual depths) to ensure that when the deviation between predicted and actual depths is within the allowable range (e.g., ≤5%), the confidence decay does not exceed 20%.

[0061] The design principle of this confidence index calculation formula is to integrate a two-dimensional evaluation of "material probability matching degree" and "reasonableness of depth deviation", where the first dimension is the material probability matching degree. The maximum probability value output by the second prediction sub-model is directly used to characterize the statistical matching degree between the predicted material type and the dielectric constant. The closer this value is to 1, the higher the degree of agreement between the dielectric constant and the typical value of a certain type of material, and the stronger the reliability of the material prediction. The second dimension is the reasonableness of the depth deviation. The impact of the deviation between the predicted depth and the reference depth is quantified by an exponential function, i.e., when the predicted depth z differs from the reference depth... When consistent, this item is 1, and the confidence level is not reduced; when divergent... As the value increases, this term decays exponentially; the larger the deviation, the more significant the decay, reflecting the logic that the further the deviation from the known pattern, the lower the confidence level. (Ceiling coefficient) Adjust the sensitivity of deviations to adapt to different geological stability scenarios;

[0062] The confidence index achieves a dual constraint of reliable material matching and reasonable depth deviation through the product of the two factors, avoiding the limitations of single-dimensional evaluation (such as unreliable results where the material probability is high but the depth deviates significantly from known patterns); final value The closer the value is to 1, the more consistent the prediction results are with known geological patterns in terms of both material type and depth location, and the higher the reliability.

[0063] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A data prediction method for achieving high-resolution subsurface layering in deep exploration, characterized by, Includes the following steps: Step 1: Use a signal transmitter and receiver to conduct multi-mode geological scanning of the underground to collect raw underground reflection signal data; Step 2 involves preprocessing, time-frequency transformation, and feature extraction of the collected raw underground reflection signal data to obtain time-domain and frequency-domain feature parameters. The preprocessing includes, in sequence, denoising, normalization, and time-frequency transformation. The denoising employs a wavelet transform algorithm, and the normalization uses a min-max normalization method. The time-frequency transformation uses a short-time Fourier transform, employing a sliding window function to transform the original time-domain reflection signal. The signal is divided into several short-time segments, and a Fourier transform is performed on each segment to obtain a complex-valued signal in the frequency-time domain. and phase spectrum Then, based on the propagation law of electromagnetic waves, the time dimension t is mapped to the underground depth dimension z to obtain the frequency-depth domain complex signal. Further extraction of energy spectrum Group delay Instantaneous frequency That is, the time-domain and frequency-domain characteristic parameters; Step 3: Input the extracted time-domain and frequency-domain feature parameters into the pre-built coupled prediction model for sequential calculation, and output the prediction results of the depth range and material type of the subsurface layer; The coupled prediction model consists of a first prediction sub-model and a second prediction sub-model. The first prediction sub-model is used to calculate the dielectric constant distribution at different depths underground based on the time-domain and frequency-domain feature parameters extracted in step two. The first prediction sub-model is as follows: ； in, Let z be the dielectric constant distribution, representing the relative dielectric constant at a depth z below ground. These are the model gain coefficients, obtained through calibration using known geological samples; The attenuation coefficient is determined based on the geological characteristics of the detection area; The frequency decay exponent was obtained by fitting experimental data. The second prediction sub-model is used to predict the dielectric constant distribution output by the first prediction sub-model. The second prediction sub-model is used to predict the material type at different depths underground. ； in, This indicates that the relative permittivity at the measured depth z is... In this case, the medium at that depth belongs to the i-th type of material. The probability of is in the range of [0,1]. This represents the predefined i-th type of underground material; The mean dielectric constant of the i-th type of material is obtained by statistical analysis of dielectric constant data of this type of material in a known geological database. The standard deviation of the dielectric constant of the i-th type of material is calculated from the dielectric constant data of this type of material in a known geological database. Step four involves comparing and verifying the prediction results obtained in step three with data from known geological databases to generate the final report on underground layered structure and material composition.

2. The data prediction method for achieving high-resolution subsurface stratification in deep earth exploration according to claim 1, characterized in that, In step one, the multi-mode geological scanning includes at least one of the following scanning methods: wide-angle reflection and refraction scanning (WARR-Scan), profile scanning (Profile-Scan), static staring scanning (Stear-Scan), and common midpoint scanning (CMP-Scan); the signal transmitter transmits pulsed radio waves and microwave beam signals to the ground through a transmitting antenna, and the group velocity of the pulsed radio waves and microwave beam signals can be adjusted by the frequency adjustment module built into the transmitter; The receiver receives the reflected signal after it has been acted upon by underground materials through a receiving antenna.

3. The data prediction method for achieving high-resolution subsurface stratification in deep earth exploration according to claim 1, characterized in that, The mapping logic that maps the time dimension t to the underground depth dimension z is as follows: Obtain the theoretical average dielectric constant of the local stratigraphic layers in the detection area. As an initial approximation, it is based on the initial value of the propagation speed of electromagnetic waves in the underground medium. ( (The speed of light in a vacuum), combined with the time relationship of two-way signal propagation. To obtain the initial depth mapping relationship Thus Converted into an initial frequency-depth domain complex signal Its energy spectrum is ,Right now , which represents the energy intensity of the reflected signal at a depth z and a frequency f.

4. The data prediction method for achieving high-resolution subsurface stratification in deep earth exploration according to claim 3, characterized in that, The group delay Instantaneous frequency The method for determining it is as follows: ； in, This represents the phase spectrum obtained after Fourier transforming the original underground reflected signal. The time-domain expression representing the original underground reflected signal. Represents the time-domain reflected signal Take the argument angle, Indicates signal frequency. Indicates time.

5. The data prediction method for achieving high-resolution subsurface stratification in deep earth exploration according to claim 1, characterized in that, In step three, the process of outputting the predicted depth range and material type of the underground strata is as follows: Material type probability distribution based on the output of the second prediction sub-model The location where the predicted material type changes at adjacent depths is taken as the stratification interface. The depth range between the interfaces is the depth range of a subsurface stratification, and each stratification corresponds to a unique predicted material type.

6. The data prediction method for achieving high-resolution subsurface stratification in deep earth exploration according to claim 5, characterized in that, In step four, the comparison and verification process includes: The subsurface stratification depth range and predicted material type of each stratum determined in step three are matched with standard stratigraphic data of the same or similar areas in the known geological database, and the confidence index of the prediction result of each stratum is calculated. The final generated report on underground layered structure and material composition includes a visualized underground layered profile, and labels the predicted material type, depth range and corresponding confidence index for each layer; The confidence index is calculated using the following formula: ； in, This represents the confidence level of the prediction result at a subsurface depth z, with a value range of [0,1]. This represents the maximum probability among all material types at depth z. This indicates the reference depth of the corresponding stratum in a known geological database; This represents the confidence attenuation coefficient, which is determined based on the geological stability of the detection area. The overall confidence level for each subsurface stratum is the sum of the values ​​at all z locations within the depth range of that stratum. The average value.

Images