Single-hole drilling radar tunnel geological advanced directional detection method

Abstract

The application discloses a single-hole drilling radar tunnel geological advanced directional detection method, comprising the following steps: step one: according to a single-hole drilling radar device, a corresponding observation system is designed, and drilling radar data is collected; step two: radar wave records RE in an advanced direction are extracted; step three: a time delay transformation method is proposed, and the radar wave records RE are transformed into equal-delay radar records RET; step four: a focusing reverse-time migration imaging method is proposed, cross-correlation calculation is performed on wave field snapshot data of forward continuation and reverse-time continuation, and equal-delay reverse-time migration data is obtained; step five: the process of step four is repeated, and a plurality of groups of equal-delay reverse-time migration data are obtained by respectively setting different equal delays; and step six: a directional information extraction method based on a deep belief network is proposed, a weight matrix containing advanced direction poor geological body information is extracted, and poor geological body information is obtained. The application aims at focusing on reducing the imaging range of a target body and realizing directional advanced detection of karst and other poor geological bodies.

Description

Technical Field

[0001] This invention relates to the field of engineering surveying technology, and in particular to a method for advanced directional geological detection of tunnels using single-hole drill radar. Background Technology

[0002] Ground-penetrating radar (GPR) boasts advantages such as being non-destructive, efficient, and high-resolution; however, it faces challenges like severe signal attenuation and limited resolution during deep exploration. Especially under complex geological conditions, such as karst areas and fractured zones, the detection performance of GPR is often unsatisfactory. The emergence of single-hole radar (SNR) technology offers a new approach to solving these problems. By placing an antenna within the borehole, SNR can directly acquire the electromagnetic response of the surrounding medium, avoiding interference from surface clutter and significantly improving the resolution and signal-to-noise ratio of deep exploration. This technology has demonstrated significant application value in fields such as mineral resource exploration, engineering geological surveys, and underground cavity detection.

[0003] The development of single-aperture radar (SAP) technology can be traced back to the 1980s. Early SAP systems primarily used low-frequency antennas (<100MHz), sacrificing resolution for greater detection depth. With advancements in electronics, modern SAP systems can now operate over a wide bandwidth of 50 MHz to 1 GHz, improving resolution while maintaining sufficient detection depth. In recent years, the rapid development of computing technology has also led to significant progress in SAP data processing algorithms. From simple filtering to depth migration imaging, and from manual interpretation to machine learning-based automatic identification, innovations in data processing technology have greatly enhanced the detection accuracy and application range of SAP.

[0004] In special scenarios such as tunnels and pile foundations, single-hole drilling radar requires directional detection of targets ahead of the borehole, and data processing requires focused imaging. Existing methods have not solved these problems. Combining the special observation system and electromagnetic wave propagation principle of single-hole drilling radar, this invention proposes a single-hole drilling radar method for advanced directional geological detection in tunnels, aiming to focus the detection range of the target and achieve directional advanced detection of adverse geological bodies such as karst. Summary of the Invention

[0005] In view of at least one deficiency of the prior art, the purpose of this invention is to provide a single-hole drill radar method for advanced directional geological detection in tunnels, which aims to focus and narrow the imaging range of the target body, realize precise directional advanced detection of adverse geological bodies such as karst, and provide technical support for tunnel excavation and pile foundation safety assessment.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A method for advanced directional geological detection of tunnels using single-hole drill radar includes the following steps: Step 1: Based on the single-hole drilling radar equipment, design an observation system with single-point transmission and single-point reception, single-point transmission and multi-point reception, or multi-point transmission and multi-point reception to collect drilling radar data; Step 2: Process the single-hole drilling radar data according to the designed observation system and extract the radar wave record (RE) in the leading direction; Step 3: Propose a time delay transformation method, and assign different time delays Δ to each radar wave record RE. t It is transformed into equal-delay radar record (RET).

[0007] Step 4: A focused reverse time migration imaging method is proposed. The reverse time extension calculation is performed using the radar record RET with equal delay as the back propagation wave field, and the forward extension calculation is performed with a wavelet of the same equal delay. The wave field snapshot data of the forward extension and the reverse time extension are cross-correlated to obtain the equal delay reverse time migration data. Step 5: Repeat the process of Step 4, and obtain multiple sets of counter-time offset data with equal delays by giving different equal delays respectively; Step 6: Propose a directional information extraction method based on deep belief networks. Using reverse time migration data with different time delays as input data, extract a weight matrix containing information on adverse geological bodies in the leading direction, and finally extract the information on adverse geological bodies in the leading direction.

[0008] The specific details of designing multiple observation systems in step one are as follows: ① Single-point transmission and single-point reception The borehole radar equipment has only one transmitting unit and one receiving unit. The transmitting and receiving units are set up independently or are integrated into one structure. The transmitting and receiving units collect zero-offset data or fixed-offset data during the movement of the borehole radar probe.

[0009] ② Single-point transmission and multi-point reception The borehole radar equipment has one transmitting unit and multiple receiving units. The borehole radar probe collects single-shot radar data during its movement.

[0010] ③Multi-point transmission and multi-point reception The borehole radar equipment has multiple transmitting units and multiple receiving units. During the movement of the borehole radar probe, the multiple transmitting units and multiple receiving units alternately transmit and receive data to collect single-shot data, zero-offset data and fixed-offset data.

[0011] Step two involves extracting radar wave records in the leading direction, including: Single-shot data, zero-offset data, and fixed-offset data contain radar wavefields in both the leading and trailing directions. The leading direction is the direction consistent with tunnel excavation, while the trailing direction is the direction opposite to tunnel excavation. The primary targets detected by borehole radar are adverse geological formations in the leading direction. This invention proposes using...FK A transformation method is used to separate the wavefield in the leading and trailing directions. First, the radar data is transformed from the time domain to... FK The domain, based on the positive and negative wavenumbers FK The spectrum is divided into positive and negative spectra, which are then transformed into the time domain to obtain left-traveling and right-traveling waves, i.e., negative wavenumber and positive wavenumber. Based on the observation system arrangement, the left-traveling wave originates from the leading direction wave field.

[0012] The specific calculation formula for extracting radar wave records in the leading direction is: (1); For a left-traveling wave field, For right-traveling wave field, For the overall wave field; specifically, the Fourier transform method is applied to extract the single-shot data. Transform from the time domain to the FK domain; in formula (1), t It is time. x It refers to the horizontal position; (2); It is the two-dimensional spectrum obtained after Fourier transform. f It's frequency. k Wavenumbers can be divided into positive wavenumbers and negative wavenumbers: (3); In the formula It is a negative wavenumber. It is the positive wave number, the spectrum. Can be divided into (4); Specifically, the negative wavenumber spectrum is preserved. And the positive wavenumber spectrum Setting it to zero, the radar wave in the leading direction can be obtained through inverse Fourier transform. ; (5); That is, the radar wave field in the forward direction.

[0013] The specific details of the time delay transformation method in step three include: assigning different time delays Δ to each radar wave record RE according to the observation system arrangement. t After transformation, the equal-delay radar record RET is obtained.

[0014] In step four, the wavelet delay transformation during the forward continuation process of focusing on reverse-time migration includes the following: During the forward and reverse-time continuation calculations of reverse-time migration, the transformed wavelet and radar wave record RET are used as the input wavefields to begin iterative calculations; the reverse-time migration imaging results are calculated using the cross-correlation imaging conditions, and the calculation formula is as follows: (6); S ( x , z , t () are different times t In spatial location ( x , z Positive extended wave field at () location; R ( x , z , t () are different times t The inverse time-delayed topological wavefield at spatial location (x, z); RESULT( x , z ) represents the reverse time offset result, where x and z are the horizontal and vertical positions in two-dimensional space, respectively.

[0015] Step 5 provides different time delays Obtain the corresponding reverse time offset data , (7); i Different time delays are used to precisely extract information about subsequent adverse geological bodies. i The value ranges from 1 to 30.

[0016] Step six of the deep belief network method for extracting information on adverse geological bodies in the leading direction includes the following specific steps: by The input layer has a size of m × n × p, where m is the number of sampling points per channel, n is the number of channels, and p is the total amount of reverse-time offset data after delay transformation. The specific algorithm flow is as follows: ① Train a deep belief network model with a small amount of input layer data and extract the weight matrix function. W The weight matrix function W It is a set of sub-weight matrices, containing m²×n² sub-weight matrices, each of size m¹×n¹; using W 11 , W 12 , …, W m2n2 Represents the sub-weight matrix; ② Each sub-weight matrix is ​​convolved with the input data to obtain m²×n² convolution results of size (m-m1+1)×(n-n1+1); when the input layer contains p-order data, m²×n²×p convolution results of size (m-m1+1)×(n-n1+1) can be obtained. R 11 , R 12 ,…, R m2n2 The result of the convolution; (8); i = 1, 2, …, p; ③ Finally, a linear weighted stacking method is applied to stack the convolution results p times to obtain the final stacked result of the reverse time-shifted data. S 11 , S 12 , …, S m2n2 This is the result of multiple superpositions of the final reverse-time offset data; (9).

[0017] Different sub-weight matrix functions can indicate different features of the true target layer. The extracted sub-weight matrix is ​​equivalent to a filter. By performing a convolution operation with the input layer data, it can extract the leading-direction reflected wave separately, while also filtering out random noise, thus achieving fine extraction of the leading-direction reflected wave and improving the resolution of the reflected wave.

[0018] Significant effects: This invention provides a single-hole drill radar method for advanced directional geological detection in tunnels, which can focus and narrow the imaging range of the target, enabling precise directional advanced detection of adverse geological bodies such as karst, and providing technical support for tunnel excavation and pile foundation safety assessment. Attached Figure Description

[0019] Figure 1 A schematic diagram of the delay transformation of this invention; Figure 2 A schematic diagram illustrating the computational flow of the reflected wave extraction method based on deep belief networks; Figure 3 The relative permittivity model shows the positions of the transmitting and receiving antennas in red. Figure 4 Rack wavelets with different phase delays: (a): 0; (b): 2; (c): 4; (d): 6; (e): 8; (f): 10; (g): 12; (h): 14; (i): 16; Figure 5Wavefield snapshot at 10ms: (a)-(i) correspond to wavelets with different full-column phases; Figure 6 Wavefield snapshot at 20ms: (a)-(i) Wavelets corresponding to different full-column phases; Figure 7 Self-excitation and self-absorption profiles simulated by wavelet of different phases; Figure 8 Comparison of curves at a depth of 0.8 m; Figure 9 Comparison of offset imaging: (a) conventional single-point emission reverse time offset method; (b) equal-delay focusing reverse time offset method; Figure 10 : Flowchart of the method of the present invention. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0021] This invention relates to a single-hole drill radar method for advanced directional geological detection in tunnels, which aims to focus and narrow the imaging range of the target body, achieve precise directional advanced detection of adverse geological bodies such as karst, and provide technical support for tunnel excavation and pile foundation safety assessment.

[0022] This invention is applicable to directional detection in scenarios such as tunnel geology in the advance direction and karst at the bottom of pile foundations.

[0023] like Figures 1-10 As shown, in order to achieve the above objectives, the present invention adopts the following technical solution: A method for advanced directional geological detection of tunnels using single-hole drill radar includes the following steps: Step 1: Based on the single-hole drilling radar equipment, design an observation system with single-point transmission and single-point reception, single-point transmission and multi-point reception, or multi-point transmission and multi-point reception to collect drilling radar data; Step 2: Process the single-hole drilling radar data according to the designed observation system and extract the radar wave record (RE) in the leading direction; Step 3: Propose a time delay transformation method, and assign different time delays Δ to each radar wave record RE. t It is transformed into equal-delay radar record (RET).

[0024] Step 4: A focused reverse time migration imaging method is proposed. The reverse time extension calculation is performed using the radar record RET with equal delay as the back propagation wave field, and the forward extension calculation is performed with a wavelet of the same equal delay. The wave field snapshot data of the forward extension and the reverse time extension are cross-correlated to obtain the equal delay reverse time migration data. Step 5: Repeat the process of Step 4, and obtain multiple sets of counter-time offset data with equal delays by giving different equal delays respectively; Step 6: Propose a directional information extraction method based on deep belief networks. Using reverse time migration data with different time delays as input data, extract a weight matrix containing information on adverse geological bodies in the leading direction, and finally extract the information on adverse geological bodies in the leading direction.

[0025] The specific details of designing multiple observation systems in step one are as follows: ① Single-point transmission and single-point reception The borehole radar equipment has only one transmitting unit and one receiving unit. The transmitting and receiving units are set up independently or are integrated into one structure. The transmitting and receiving units collect zero-offset data or fixed-offset data during the movement of the borehole radar probe.

[0026] ② Single-point transmission and multi-point reception The borehole radar equipment has one transmitting unit and multiple receiving units. The borehole radar probe collects single-shot radar data during its movement.

[0027] ③Multi-point transmission and multi-point reception The borehole radar equipment has multiple transmitting and receiving units. During the movement of the borehole radar probe, multiple transmitting and receiving units alternately transmit and receive data, collecting single-shot data, zero-offset data, and fixed-offset data.

[0028] Step two involves extracting radar wave records in the leading direction, including: Single-shot data, zero-offset data, and fixed-offset data contain radar wavefields in both the leading and trailing directions. The leading direction is the direction consistent with tunnel excavation, while the trailing direction is the direction opposite to tunnel excavation. The primary targets detected by borehole radar are adverse geological formations in the leading direction. This invention proposes using... FK A transformation method is used to separate the wavefield in the leading and trailing directions. First, the radar data is transformed from the time domain to... FK The domain, based on the positive and negative wavenumbers FK The spectrum is divided into positive and negative spectra, which are then transformed into the time domain to obtain left-traveling waves (negative wavenumber) and right-traveling waves (positive wavenumber). Depending on the arrangement of the observation system, the left-traveling wave usually originates from the leading direction wave field.

[0029] The specific calculation formula for extracting radar wave records in the leading direction is: (1); For a left-traveling wave field, For right-traveling wave field, For the overall wave field; specifically, the Fourier transform method is applied to extract the single-shot data. Transform from the time domain to the FK domain; in formula (1), t It is time.x It refers to the horizontal position; (2); It is the two-dimensional spectrum obtained after Fourier transform. f It's frequency. k Wavenumbers can be divided into positive wavenumbers and negative wavenumbers: (3); In the formula It is a negative wavenumber. It is the positive wave number, the spectrum. Can be divided into (4); Specifically, the negative wavenumber spectrum is preserved. And the positive wavenumber spectrum Setting it to zero, the radar wave in the leading direction can be obtained through inverse Fourier transform. ; (5); That is, the radar wave field in the forward direction.

[0030] The specific details of the time delay transformation method in step three are as follows: Figure 1 As shown, for each radar wave record RE, a different time delay Δ is given according to the arrangement of the observation system. t After transformation, the equal-delay radar record RET is obtained.

[0031] In step four, the wavelet delay transform during the forward extension process in the focused reverse-time offset is similar to... Figure 1 In the forward and reverse time continuation calculations of reverse time migration, the transformed wavelet and radar wave record RET are used as the input wavefields to begin iterative calculations; the reverse time migration imaging results are obtained by applying the cross-correlation imaging conditions, and the calculation formula is as follows: (6); S ( x , z , t () are different times t In spatial location ( x , z Positive extended wave field at () location; R ( x , z , t () are different times t The inverse time-delayed topological wavefield at spatial location (x, z); RESULT( x , z The result is the reverse time offset.x and z These represent the horizontal and vertical positions in two-dimensional space, respectively.

[0032] Step 5 provides different time delays Obtain the corresponding reverse time offset data , (7); i Different time delays are used to precisely extract information about subsequent adverse geological bodies. i The value ranges from 1 to 30.

[0033] The specific steps for extracting information on adverse geological bodies in the leading direction using the deep belief network method in step six are as follows: Figure 2 As shown. The input layer has a size of m × n × p, where m is the number of sampling points per channel, n is the number of channels, and p is the total amount of reverse-time offset data after delay transformation. The specific algorithm flow is as follows: ① Train a deep belief network model with a small amount of input layer data and extract the weight matrix function. W The weight matrix function W It is a set of sub-weight matrices, containing m²×n² sub-weight matrices, each of size m¹×n¹; using W 11 , W 12 , …, W m2n2 Represents the sub-weight matrix; ② Each sub-weight matrix is ​​convolved with the input data to obtain m²×n² convolution results of size (m-m1+1)×(n-n1+1). When the input layer contains p-th iterations of data, m²×n²×p convolution results of size (m-m1+1)×(n-n1+1) can be obtained. R 11 , R 12 ,…, R m2n2 This is the result of the convolution; (8); i = 1, 2, …, p.

[0034] ③ Finally, a linear weighted superposition method is applied to superimpose the results of p convolutions to obtain the final superposition result. S 11 , S 12 , …, S m2n2 This is the result of multiple superpositions.

[0035] (9); Different sub-weight matrix functions can indicate different features of the true target layer. The extracted sub-weight matrix is ​​equivalent to a filter. By performing a convolution operation with the input layer data, it can extract the leading-direction reflected wave separately, while also filtering out random noise, thus achieving fine extraction of the leading-direction reflected wave and improving the resolution of the reflected wave.

[0036] The specific advantages of this invention are: (1) The time delay transformation method does not lose any wave field information and can completely preserve all the characteristics of the radar wave field; (2) The focused reverse time migration method can flexibly call radar wave records with different equal delays to implement high-precision migration imaging; (3) The information extraction method based on deep belief network can extract a high-precision weight matrix using only a small sample training data, and obtain detailed information on adverse geological bodies in the leading direction.

[0037] This invention creatively proposes a single-hole drilled radar method for advanced directional geological detection in tunnels. By using a time delay transformation method to correct the radar wave field into radar wave fields with different equal delays, and a reverse time migration method to provide high-precision imaging, the invention utilizes a deep belief network information extraction method to obtain detailed information on adverse geological bodies in the advanced direction. This enables precise directional detection in the advanced direction and provides technical support for tunnel excavation and pile foundation safety assessment.

[0038] The single-hole drilling radar method for advanced directional geological detection and focused imaging in tunnels falls within the scope of patent protection. Its application scenarios are not limited to tunnels as described in the patent specification; pile foundation applications also fall within the scope of patent protection.

[0039] Model Experiment To verify the method of this invention, a layered model was designed, such as... Figure 3 As shown, the relative permittivity is 2 and 4, respectively. The sampling interval is set to 0.05 ns, and the number of sampling points per channel is 2000. The red circles represent the transmitting and receiving units, totaling 12. The transmitting and receiving units transmit and receive radar waves simultaneously. Given equal-delay radar wavelets with intervals of 2 sampling points from 0 to 16, as shown... Figure 4 As shown, the Reichschild wavelets with different phase delays are (a): 0; (b): 2; (c): 4; (d): 6; (e): 8; (f): 10; (g): 12; (h): 14; (i): 16; and wave field snapshots at different times are obtained by simulation. Figure 5 and Figure 6Wavefield snapshots at 10 ns and 20 ns are shown, with A, B, and C as wavefronts. In advance detection, the aim is to minimize the energies of wavefronts A and C, and concentrate the propagation energy as much as possible at B. (Comparison) Figure 5 and Figure 6 Wavefront energies with different delays Figure 5 (i) and Figure 6 (i) The energy of wavefront B is most concentrated in the middle wave. Figure 7 The images show self-excited and self-receiving radar profiles obtained from simulations using wavelets with different delay times. Higher frequencies in the waveform profiles indicate stronger target resolution. Figure 7 The waveform resolution in (a) is relatively low. Figure 7 (a) to Figure 7 (i) The resolution gradually increases, among which Figure 7 (i) has the highest radar wave resolution. Figure 8 The curve is displayed at a depth of 0.8 m, where Figure 8 (i) The (red) curve has the highest amplitude energy, which also verifies that... Figure 8 (i) The energy of the mid-wave field is most focused. Figure 9 The images show the migration profiles obtained using the conventional single-point emission reverse-time migration method (a) and the equal-delay focusing reverse-time migration method (b), respectively. A, B, and C represent interfering reflected waves, and D represents the actual reflection interface. During advance detection, the energy is strongest at the desired region D, and lowest at locations A, B, and C. Figure 9 In (a), the energy differences between B, C, and D are relatively small, mainly because the conventional reverse-time migration method cannot achieve energy focusing; the electromagnetic wave energy is uniformly distributed at positions B, C, and D. (Comparison) Figure 9 (a), Figure 9 (b) The energy at D is the most concentrated, while the energy at B and C is the lowest, resulting in the highest resolution at the interface. The energy at A is also enhanced, but this does not affect the resolution at the interface D.

[0040] Numerical simulation results demonstrate that, compared to traditional single-hole drilling radar detection methods, the method provided by this invention changes the propagation direction of electromagnetic waves through equal delay transformation, thereby focusing the propagation energy of electromagnetic waves directly in front, which can more clearly display the structural features of the interface, suppress the interference of artifact noise, and improve the resolution of the target.

[0041] The invention is currently primarily used in the field of advanced geological forecasting for tunnels, but it can also be used to investigate defects such as karst caves at the bottom of pile foundations.

[0042] Finally, it should be noted that the above are only specific embodiments of the present invention. Of course, those skilled in the art can make modifications and variations to the present invention. If these modifications and variations fall within the scope of the claims of the present invention and their equivalents, they should be considered as being within the protection scope of the present invention.

Claims

1. A method for advanced directional geological detection of tunnels using single-hole drill radar, characterized in that, Includes the following steps: Step 1: Based on the single-hole drilling radar equipment, design an observation system with single-point transmission and single-point reception, single-point transmission and multi-point reception, or multi-point transmission and multi-point reception to collect drilling radar data; Step 2: Process the single-hole drilling radar data according to the designed observation system and extract the radar wave record (RE) in the leading direction; Step 3: Propose a time delay transformation method, and assign different time delays Δ to each radar wave record RE. t Transformed into equal-delay radar record (RET); Step 4: A focused reverse time migration imaging method is proposed. The reverse time extension calculation is performed using the radar record RET with equal delay as the back propagation wave field, and the forward extension calculation is performed with a wavelet of the same equal delay. The wave field snapshot data of the forward extension and the reverse time extension are cross-correlated to obtain the equal delay reverse time migration data. Step 5: Repeat the process of Step 4, and obtain multiple sets of counter-time offset data with equal delays by giving different equal delays respectively; Step 6: Propose a directional information extraction method based on deep belief networks. Using reverse time migration data with different time delays as input data, extract a weight matrix containing information on adverse geological bodies in the leading direction, and finally extract the information on adverse geological bodies in the leading direction.

2. The single-hole drilling radar tunnel geological advance directional detection method according to claim 1, characterized in that: Step one involves designing various observation systems, including: ① Single-point transmission and single-point reception The borehole radar equipment has only one transmitting unit and one receiving unit. The transmitting unit and receiving unit are set up independently or are integrated into one structure. The transmitting unit and receiving unit collect zero-offset data or fixed-offset data during the movement of the borehole radar probe. ② Single-point transmission and multi-point reception The drilling radar equipment has one transmitting unit and multiple receiving units. The drilling radar probe collects single-shot radar data during its movement. ③Multi-point transmission and multi-point reception The borehole radar equipment has multiple transmitting units and multiple receiving units. During the movement of the borehole radar probe, the multiple transmitting units and multiple receiving units alternately transmit and receive data to collect single-shot data, zero-offset data and fixed-offset data.

3. The single-hole drilling radar tunnel geological advance directional detection method according to claim 2, characterized in that: Step two involves extracting radar wave records (REs) in the leading direction, including: Single-shot data, zero-offset data, and fixed-offset data contain radar wavefields in both the leading and trailing directions; the leading direction is the direction consistent with tunnel excavation, while the trailing direction is the direction opposite to tunnel excavation; the target detected by borehole radar is an unfavorable geological body in the leading direction; [The text then abruptly shifts to a different topic:] ...using FK The transformation method achieves wavefield separation in the leading and trailing directions; first, the radar data is transformed from the time domain to... FK The domain, based on the positive and negative wavenumbers FK The spectrum is divided into positive and negative spectra, which are then transformed into the time domain to obtain left-traveling and right-traveling waves, i.e., negative wave number and positive wave number; according to the arrangement of the observation system, the left-traveling wave is the wave field from the leading direction. The specific calculation formula for extracting radar wave records in the leading direction is: (1) ； For a left-traveling wave field, For right-traveling wave field, This represents the overall wavefield; specifically, it involves applying the Fourier transform method to analyze the single-shot data. Transform from the time domain to the FK domain; (2) ； It is the two-dimensional spectrum obtained after Fourier transform. f It's frequency. k Wavenumbers can be divided into positive wavenumbers and negative wavenumbers: (3) ； In the formula It is a negative wavenumber. It is the positive wave number, the spectrum. Can be divided into (4) ； Specifically, the negative wavenumber spectrum is preserved. And the positive wavenumber spectrum Setting it to zero, the radar wave in the leading direction can be obtained through inverse Fourier transform. ; (5) ； That is, the radar wave field in the forward direction.

4. The single-hole drilling radar tunnel geological advance directional detection method according to claim 1, characterized in that: In step four, the wavelet delay transformation during the forward extension process in the focused reverse-time offset includes: In the forward and reverse time continuation calculations of reverse time migration, the transformed wavelet and radar wave record RET are used as the input wavefields to begin iterative calculations. The reverse time migration imaging results are obtained by applying the cross-correlation imaging conditions, and the calculation formula is as follows: （6）； S ( x , z , t () are different times t In spatial location ( x , z Positive extension wave field at () location; R ( x , z , t () are different times t The inverse time-delayed topological wavefield at spatial location (x, z); RESULT( x , z The result is the reverse time offset. x and z These represent the horizontal and vertical positions in two-dimensional space, respectively.

5. The single-hole drill radar tunnel geological advance directional detection method according to claim 1, characterized in that, Step 5 provides different time delays Obtain the corresponding reverse time offset data , （7）； i Different time delays are used to precisely extract information about subsequent adverse geological bodies. i The value ranges from 1 to 30.

6. The single-hole drilling radar tunnel geological advanced directional detection method according to claim 5, characterized in that, The specific steps for extracting information on adverse geological bodies in the leading direction using the deep belief network method in step six include: by The input layer has a size of m × n × p, where m is the number of sampling points per channel, n is the number of channels, and p is the total amount of reverse-time offset data after delay transformation. The specific algorithm flow is as follows: ① Train a deep belief network model with a small amount of input layer data and extract the weight matrix function. W The weight matrix function W It is a set of sub-weight matrices, containing m²×n² sub-weight matrices, each of size m¹×n¹; using W 11 , W 12 , …, W m2n2 Represents the sub-weight matrix; ② Each sub-weight matrix is ​​convolved with the input data to obtain m²×n² convolution results of size (m-m1+1)×(n-n1+1); when the input layer contains p-order data, m²×n²×p convolution results of size (m-m1+1)×(n-n1+1) can be obtained. R 11 , R 12 ,…, R m2n2 The result of the convolution; (8) ； i = 1, 2, …, p； ③ Finally, a linear weighted stacking method is applied to stack the convolution results p times to obtain the final stacked result of the reverse time-shifted data. S 11 , S 12 , …, S m2n2 This is the result of multiple superpositions of the final reverse-time offset data; (9)。

Images