Underground in-situ detection method and system suitable for kilometer level horizontal hole

Abstract

The invention relates to the field of geological detection, provides an underground in-situ detection method and system suitable for a kilometer-level horizontal hole in order to improve the detection efficiency and precision at the same time, constructs an evidence reliability guarantee mechanism of hardware noise reduction and algorithm weighting, dynamically adjusts the weights of earthquake and radar data according to the quality fluctuation of data while drilling, and improves the detection efficiency and precision. The reliability of fusion input data is ensured; a D-S algorithm redundancy calculation step is simplified, a fusion calculation period is compressed to a millisecond level in combination with a high-speed data transmission link, and the real-time feedback requirement while drilling is met; a three-dimensional collaborative detection and fusion system of a macroscopic geologic structure and a microscopic anomalous body is constructed, the advantages of multi-source data are brought into full play, and the comprehensiveness and the refinement degree of underground geologic structure detection are improved.

Description

Technical Field

[0001] This invention relates to the field of geological exploration, specifically to an in-situ underground exploration method and system applicable to kilometer-scale horizontal boreholes. Background Technology

[0002] In the field of geological engineering, accurately obtaining information on underground geological structures is a core prerequisite for resource exploration, engineering design, and construction. With the advancement of national deep mineral resource development and large-scale underground space construction, the demand for "exploration efficiency, accuracy, and comprehensiveness" in-situ exploration using kilometer-level horizontal boreholes is becoming increasingly urgent. Currently, underground in-situ exploration technologies are mainly divided into two categories: drilling and geophysical exploration. The integrated exploration mode combining these two has become an industry trend, but key technical bottlenecks still exist at the data fusion level.

[0003] From a technological perspective, existing integrated exploration equipment mostly adopts a simple hardware overlay model of "drilling + geophysical exploration," and has not yet formed a mature multi-source data collaborative fusion system. On the one hand, seismic exploration and borehole radar exploration, as two core geophysical exploration methods, each have their own data limitations: seismic data is good at reflecting macroscopic geological structures (such as faults and karst caves), but its ability to identify microscopic anomalies (such as fissures and aquifers) is insufficient; radar data can achieve local high-resolution imaging, but its detection range is limited and it is easily affected by geological environment. Single data has the problems of "multiple interpretations" and "information bias," making it difficult to form a comprehensive and accurate geological judgment. On the other hand, multi-source data fusion technology is not well adapted to drilling exploration scenarios: existing fusion methods (such as simple weighted average and logical judgment methods) do not fully consider the dynamic characteristics of drilling data (such as signal noise caused by drilling vibration and real-time fluctuations in data quality), and cannot effectively quantify the reliability of different data sources; while DS evidence theory, as a classic method of multi-source information fusion, has been initially applied in some static geological exploration scenarios (such as tunnel water inrush prediction), it faces three core adaptation challenges when directly transferred to the kilometer-level horizontal hole drilling scenario: First, environmental factors such as vibration and high mineralization during drilling cause high data noise, making it difficult to guarantee the credibility of evidence; second, drilling exploration requires millisecond-level real-time feedback to guide drilling, and conventional DS algorithms are computationally cumbersome and time-consuming, which cannot meet the real-time requirements; third, the dimensional differences between seismic and radar data are large, and the evidence differentiation is insufficient when fusing dual-source data, which can easily lead to deviations in the fusion results. Summary of the Invention

[0004] To simultaneously improve detection efficiency and accuracy, this invention provides an in-situ underground detection method and system suitable for kilometer-scale horizontal boreholes.

[0005] The technical solution adopted by the present invention to solve the above problems is:

[0006] In-situ subsurface exploration methods applicable to kilometer-scale horizontal boreholes include:

[0007] Step 1: Use a data acquisition unit with hardware-based noise reduction to acquire seismic and radar data while drilling;

[0008] Step 2: Obtain characteristic parameters based on the acquired seismic and radar data. The characteristic parameters obtained from the seismic data include at least the seismic wave signal-to-noise ratio. The characteristic parameters obtained based on radar data include at least the radar wave polarization purity P;

[0009] Step 3: Based on the seismic wave signal-to-noise ratio The reliability of the data is determined by the radar wave polarization purity P and the preset reliability classification rules.

[0010] Step 4: Determine the weights of seismic data and radar data based on data reliability. and ,and ;

[0011] Step 5: Define the recognition framework And obtain its power set, which includes simple subset propositions and composite subset propositions. ,in, For macroscopic construction, It is a microscopic anomaly. Strong vertical polarization reflection = water-bearing fracture. There are no significant geological anomalies;

[0012] Step 6: Perform feature proposition matching on the obtained feature parameters to obtain the confidence level of each proposition. Delete composite subset propositions based on the confidence level. Distribute the confidence levels of the remaining composite subset propositions to the corresponding single subset propositions according to a fixed ratio based on the engineering risk.

[0013] Step 7: Based on the weights of seismic data and radar data and Adjust the confidence level of each subset proposition: , , The confidence level corresponding to earthquake data. The level of trust in the radar data. , The revised trust level;

[0014] Step 8: Calculate the fusion confidence of all propositions using the DS combination rule based on the revised confidence level to determine the detection results.

[0015] Furthermore, the hardware noise reduction methods include: setting up shock-absorbing rubber pads, a suspension shock-absorbing structure based on metal springs and dampers, and / or a vibration isolation cover.

[0016] Furthermore, the preset reliability classification rules are as follows: For high reliability, For reliability, For low reliability, The high reliability threshold corresponding to the signal-to-noise ratio of seismic waves. This represents the low reliability threshold corresponding to the signal-to-noise ratio of seismic waves.

[0017] For high reliability, For reliability, For low reliability, This represents the high reliability threshold corresponding to the polarization purity of radar waves. This represents the low reliability threshold corresponding to the polarization purity of radar waves.

[0018] Furthermore, if and Then, earthquake and radar data will be collected again;

[0019] like and ,but , ;

[0020] like and ,but , ;

[0021] Otherwise, first consider the signal-to-noise ratio of seismic waves. The reliability of radar wave polarization purity P is determined by seismic and radar data.

[0022] Basic weights and Then, based on the basic weights and calculate and : , .

[0023] Furthermore, basic weights and Specifically:

[0024] when and hour, , ;

[0025] when and hour, , ;

[0026] when and hour, , ;

[0027] when and hour, , .

[0028] Furthermore, the high reliability threshold for the signal-to-noise ratio of seismic waves is 30 dB, and the low reliability threshold is 15 dB; the high reliability threshold for the polarization purity of radar waves is 90%, and the low reliability threshold is 80%.

[0029] Furthermore, the composite subset proposition retains only { }、{ }

[0030] Furthermore, it also includes: Step 9, calculating the confidence interval based on the fused confidence value to determine the reliability level of the fusion result.

[0031] Furthermore, it also includes: Step 10, making engineering decisions based on the fusion results and reliability levels.

[0032] Subsurface in-situ detection systems suitable for kilometer-scale horizontal boreholes include:

[0033] Data acquisition unit: used for acquiring seismic and radar data while drilling;

[0034] Data processing unit: Processes the data collected by the data acquisition unit based on the underground in-situ detection method applicable to kilometer-level horizontal boreholes.

[0035] The advantages of this invention compared to existing technologies are as follows: It constructs an evidence reliability assurance mechanism of "hardware noise reduction + algorithm weighting," dynamically adjusting the weights of seismic and radar data to address fluctuations in drilling data quality and ensure the reliability of the fused input data; it simplifies redundant calculation steps in the DS algorithm, compressing the fusion calculation cycle to milliseconds by combining a high-speed data transmission link, meeting the real-time feedback requirements of drilling; and it constructs a three-dimensional collaborative detection and fusion system of "macro-geological structure + micro-anomalies," fully leveraging the advantages of multi-source data and improving the comprehensiveness and precision of underground geological structure detection. Attached Figure Description

[0036] Figure 1 This is an in-situ underground exploration method applicable to kilometer-scale horizontal boreholes;

[0037] Figure 2 This is a schematic diagram of the evidence interval in evidence theory;

[0038] Figure 3This is a diagram of the integrated platform architecture for in-situ underground exploration while drilling in kilometer-scale horizontal boreholes. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0040] In existing technologies, the closest solution to this invention is "hardware-integrated drilling detection equipment + conventional data fusion method." However, such solutions do not provide customized optimization of the DS evidence theory for drilling scenarios, resulting in problems such as low reliability of fusion results, poor real-time performance, and insufficient adaptability. They cannot fully leverage the synergistic advantages of multi-source data and are insufficient to meet the needs of refined detection in kilometer-scale horizontal boreholes. Therefore, this invention provides a subsurface in-situ detection method and system suitable for kilometer-scale horizontal boreholes.

[0041] like Figure 1 As shown, the in-situ underground exploration method applicable to kilometer-scale horizontal boreholes includes:

[0042] Step 1: Use a data acquisition unit with hardware-based noise reduction to acquire seismic and radar data while drilling.

[0043] In this embodiment, the hardware noise reduction method is as follows: a shock-absorbing rubber pad is set up. The noise amplitude caused by drilling vibration can be reduced by the shock-absorbing rubber pad, so that the signal-to-noise ratio of the seismic wave signal is increased from the previous 15dB to more than 35dB, and the polarization purity of the radar wave is maintained at more than 92%, providing hardware support for the reliability of evidence.

[0044] Step 2: Obtain characteristic parameters based on the acquired seismic and radar data. The characteristic parameters obtained from the seismic data include at least the seismic wave signal-to-noise ratio. The characteristic parameters obtained based on radar data include at least the radar wave polarization purity P.

[0045] In this embodiment, seismic data is acquired through a controllable mechanical seismic source and a seismic detector. After filtering and digital processing, parameters such as seismic wave propagation time, reflection amplitude, and seismic wave signal-to-noise ratio are extracted. Radar data is acquired through a fully polarized borehole radar antenna, and parameters such as horizontal polarization / vertical polarization amplitude, phase, and polarization purity are extracted.

[0046] Step 3: Based on the seismic wave signal-to-noise ratio The reliability of the data is determined by the radar wave polarization purity P and the preset reliability classification rules.

[0047] The core factor of seismic data is the signal-to-noise ratio of seismic waves. The signal-to-noise ratio (SNR) of seismic waves directly determines the accuracy of seismic data in identifying macroscopic geological structures (such as faults); the core factor of radar data is the radar wave polarization purity P, which directly affects the radar data's ability to distinguish microscopic geological anomalies (such as water-bearing fissures). Therefore, this invention addresses this issue by using the seismic wave SNR... The reliability of the data is determined by the polarization purity P of the radar waves.

[0048] The preset reliability classification rules are as follows: For high reliability, For reliability, For low reliability, The high reliability threshold corresponding to the signal-to-noise ratio of seismic waves. This represents the low reliability threshold corresponding to the signal-to-noise ratio of seismic waves.

[0049] For high reliability, For reliability, For low reliability, This represents the high reliability threshold corresponding to the polarization purity of radar waves. This represents the low reliability threshold corresponding to the polarization purity of radar waves.

[0050] In this embodiment, the high reliability threshold for the seismic wave signal-to-noise ratio is 30 dB, and the low reliability threshold is 15 dB; the high reliability threshold for radar wave polarization purity is 90%, and the low reliability threshold is 80%. At high reliability, noise interference is ≤10%, and polarization characteristics are undistorted; at medium reliability, noise interference is 10%-50%, and polarization characteristics are slightly distorted; at low reliability, noise interference is >50%, and polarization characteristics are severely distorted. Adaptive adjustments can be made according to actual conditions, and no restrictions are imposed here.

[0051] Step 4: Determine the weights of seismic data and radar data based on data reliability. and ,and .

[0052] like and If necessary, reacquire seismic and radar data. At this point, neither source data can reliably support fusion, requiring the triggering of a "data supplementation mechanism," without immediate weight calculation. The supplementation measures employed in this embodiment are: enhancing noise reduction through the vibration reduction module of the equipment system (e.g., activating secondary vibration reduction to reduce vibration noise by 30%), while simultaneously extending the data acquisition time (from 0.5s to 1.2s for seismic waves and from 0.3s to 0.8s for radar waves), until the data quality improves to "medium reliability" or higher. If the value is ≥15dB or P≥80%, then the weights are recalculated.

[0053] like and ,but , ;like and ,but , .

[0054] The core principle of the above formula design is:

[0055] 1. Forcefully reduce the weight of unreliable data: Use a coefficient difference of 0.1 and 0.4 to forcibly reduce the influence of low-quality data;

[0056] 2. Binding data quality dynamic adjustment: Introducing seismic data quality normalization values ​​and radar data quality normalization values, allowing the weights to fluctuate with real-time data quality, avoiding the rigidity of fixed weights;

[0057] 3. Ensure the sum of the weights is 1: The design conforms to the mathematical requirements of DS evidence theory for weights, ensuring the rationality of the fusion calculation.

[0058] by For example, a coefficient of 0.1 significantly limits the weight contribution of seismic data, even... (Low reliability threshold), the maximum value of this item is only 0.1×(15 / 15)=0.1, ensuring that the basic weight of seismic data is extremely low; The signal-to-noise ratio of seismic waves was normalized to the 0-1 range. The closer to 15dB (low reliability upper limit), the larger this value, resulting in a slight increase in the weight of seismic data, reflecting the logic of "better data quality, slightly higher weight." A coefficient of 0.4 highlights the dominant position of radar data. The polarization purity of radar waves is normalized to the 0-1 range. The higher the value of P, the larger the value of this term, and the higher the weight of radar data, ensuring the core role of highly reliable radar data.

[0059] Otherwise, first consider the signal-to-noise ratio of seismic waves. The reliability of radar wave polarization purity P determines the fundamental weights of seismic and radar data. and Then, based on the basic weights and calculate and : , .

[0060] In this embodiment, the basic weights and Set to:

[0061] when and hour, , ;

[0062] when and hour, , ;

[0063] when and hour, , ;

[0064] when and hour, , .

[0065] The setting of basic weight values ​​is essentially a quantitative definition of "the contribution of seismic / radar data of different quality levels to underground geological exploration," and the core principle is:

[0066] 1. Highly reliable data empowers equal access: local seismic wave signal-to-noise ratio When the radar wave polarization purity is ≥30dB (high reliability, noise interference ≤10%) and P ≥90% (high reliability, no distortion of polarization characteristics), both types of data can stably provide effective information. Seismic data excels at detecting macroscopic geological structures (faults, karst caves), while radar data excels at identifying microscopic anomalies (fractures, aquifers). Their detection value is complementary and their reliability is balanced; therefore, they are assigned equal weights. =0.5, =0.5), to avoid a single data point dominating the fusion result.

[0067] 2. Degradation of low-reliability data weight: When the quality of a certain type of data deteriorates (e.g., seismic data ≤ 15dB), the weight of the data is reduced. <30dB, medium reliability, noise interference 10%-50%), while when the other type of data maintains high reliability (P≥90%), the interference factors of low reliability data increase and the detection accuracy decreases, so its weight needs to be reduced; at the same time, the relative value of high reliability data increases, so its weight needs to be increased to ensure the reliability of the fusion result.

[0068] 3. Adaptation to Detection Scenarios: In kilometer-level horizontal borehole detection, macroscopic structures (requiring seismic data support) and microscopic anomalies (requiring radar data support) are both key detection targets with no absolute priority. Therefore, the weight adjustment only fluctuates within the range of "0.3-0.7", which can highlight the role of high-quality data without completely abandoning the auxiliary value of medium-reliable data.

[0069] For example, when 15dB≤ When the noise level is <30dB, the seismic wave noise increases, and the positioning error of distant macroscopic structures will expand from ±1m to ±3m, resulting in a decrease in macroscopic detection capability. When P≥90%, the radar wave polarization characteristics are undistorted, and the accuracy rate of identifying microscopic anomalies around the borehole (such as fractures ≤0.5m) remains ≥88%. Moreover, microscopic anomalies directly affect borehole safety (such as water-bearing fractures that can easily lead to borehole collapse), making real-time detection more valuable. Reducing the weight of seismic data to 0.3 (retaining only basic auxiliary role) and increasing the weight of radar data to 0.7 (dominating the fusion result) not only avoids the interference caused by seismic data noise but also enhances the guiding value of highly reliable radar data for safe drilling.

[0070] When 80%≤P<90%, the radar wave polarization characteristics are slightly distorted, and the ability to distinguish microscopic anomalies decreases (e.g., the misjudgment rate between water-bearing fissures and dry fissures increases from 5% to 15%). At a depth of ≥30dB, seismic wave propagation is stable, and the positioning error for macroscopic structures (such as faults) is still ≤±1m. However, if macroscopic structures (such as large faults) are not identified in time, the overall borehole trajectory may deviate from the target, posing a greater hazard. Reducing the weight of radar data to 0.3 (to assist in the identification of microscopic targets) and increasing the weight of seismic data to 0.7 (to dominate the detection of macroscopic structures) prioritizes ensuring the accuracy of identifying high-risk macroscopic geological targets, which is in line with the engineering logic of "avoiding major risks first, and then refining microscopic detection".

[0071] The core of this weight design is hierarchical processing:

[0072] Basic weights: Use non-normalized values ​​to define the "reliability priority" for different data quality scenarios;

[0073] Dynamic adjustment: The basic weights are transformed into normalized final weights through a formula, which satisfies the mathematical requirement of "the sum of weights equals 1" in the DS evidence theory, and at the same time achieves the core goal of "strong correlation between data quality and weights".

[0074] Step 5: Define the recognition framework And obtain its power set, which includes simple subset propositions and composite subset propositions. ,in, Macroscopic structures such as faults and karst caves, These are microscopic anomalies such as fractures / aquifers. Strong vertical polarization reflection = water-bearing fracture. There are no significant geological anomalies.

[0075] "Vertical polarization" refers to the polarization of an electromagnetic wave where the electric field direction is perpendicular to the detection reference surface (such as the borehole axis or the ground). "Strong reflection" is a direct description of the amplitude of the radar wave reflection signal; that is, when a radar wave with a vertical polarization direction encounters an underground target (such as a water-bearing fissure), the amplitude of the reflected signal is significantly higher than that of other polarization directions or background signals. The combination of these two terms, "strong vertical polarization reflection," accurately expresses the physical phenomenon of "enhanced radar reflection signal under a specific polarization direction."

[0076] In this application, the fusion of seismic data and radar data is based on the core framework of DS evidence theory and is customized to suit the characteristics of the drilling scenario.

[0077] Within the framework of DS evidence theory, if there is a problem that requires a decision, the first step is to define an identification framework for that problem. Defined as an identification framework for DS evidence theory reasoning: , by set The set of all subsets The power set called Θ is used for an identification framework containing N elements. Its power set It should include all One hypothesis: .

[0078] Suppose A is a power set Let A be an element, and let the mass function of A be m(A). Then the mapping function m satisfies the following condition: .

[0079] In DS evidence theory, the trust function and likelihood function represent the degree of confidence that proposition A is true and the degree of confidence that proposition A is not false, respectively, and are used to measure the uncertainty of evidence. The trust function is defined as follows: The likelihood function represents a measure of uncertainty about the apparent likelihood of proposition A being true (the maximum confidence that proposition A is not false), and is defined as follows: .

[0080] In the formula, the trust function Bel(A) of proposition A is equal to the sum of the basic probability assignments corresponding to all subsets of A, therefore Bel(A) represents the lower bound of the support of proposition A; while the likelihood function Pl(A) is equal to the recognition frame. The likelihood function Pl(A) is the sum of the basic probability assignments corresponding to all subsets whose intersection with proposition A is not empty. Therefore, Pl(A) represents the upper bound of the support of proposition A, and the two always satisfy the condition Bel(A) ≤ Pl(A). In fact, [Bel(A),Pl(A)] represents the uncertainty (confidence) interval of A, describing the degree of confidence in proposition A; [0,Bel(A)] represents the supporting evidence interval of proposition A, [0,Pl(A)] represents the pseudo-confidence interval of proposition A, and [Pl(A),1] represents the rejection evidence interval of proposition A; Pl(A) - Bel(A) is called the uncertainty of A, such as... Figure 2 As shown.

[0081] For multiple different pieces of evidence within the same recognition framework, i.e., multiple different quality functions, if the evidence is not completely conflicting, the DS combination rule can be used for fusion. The fusion result is usually expressed as an orthogonal sum. Assume there exists a definition within the recognition framework... Two independent mass functions and Their combination It can be calculated using the following formula:

[0082] ,

[0083] In the formula, A, B, and C all represent recognition frames. The proposition consisting of a subset of the given set, K, is called the conflict factor or conflict coefficient, used to measure the degree of conflict between pieces of evidence. The larger the K value, the higher the degree of conflict between the evidence. When K = 0, it indicates that the evidence is completely consistent. When K = 1, the evidence is completely conflicting, and the DS combination rule cannot be used for fusion.

[0084] In this embodiment, the power set of the identification framework contains 16 subsets of hypotheses, covering all possible geological conclusions in the detection scenario.

[0085] Step 6: Perform feature proposition matching on the obtained feature parameters to obtain the confidence level of each proposition, delete composite subset propositions based on the confidence level, and assign the confidence level of the remaining composite subset propositions to the corresponding single subset propositions.

[0086] After multiple pieces of evidence are integrated, the confidence level assigned to the composite subset proposition needs to be extracted again and merged into the corresponding single subset proposition in order to make a more intuitive decision.

[0087] Taking the recognition framework of this invention as an example, the original DS process requires three "tedious" steps, resulting in slow computation (typically 100ms):

[0088] 1. List all propositions concerning composite subsets.

[0089] for example: , , These are all "fuzzy propositions," and after merging, there will be a trust level allocation (e.g., m( )=0.25、m ( =0.15).

[0090] 2. Trust level of splitting composite subsets according to rules

[0091] The original rule is "split equally" or "split proportionally", for example:

[0092] Compound Proposition The trust level m = 0.25, containing 2 singlets ( , ), divided equally into: The result is 0.125. The result is 0.125;

[0093] Compound Proposition The trust level m = 0.15, containing 3 singlets ( , , ), divided equally into: The result is 0.05. The result is 0.05. The result is 0.05.

[0094] This decomposition process must be repeated for all 12 compound propositions, resulting in an extremely large amount of computation.

[0095] 3. Summarize the trust levels of all single subsets.

[0096] Add the trust scores of the split subsets to the trust scores of the original subsets, for example:

[0097] Original m ({ })=0.5, after splitting add 0.125 (from { , }),final Trust level = 0.625;

[0098] Original m ({ })=0.1, after splitting, add 0.125 (from { }) +0.05 (from { }),final Trust level = 0.275;

[0099] Similarly, calculate the final trust level for each of the four subsets. Perform a "consistency check" on the final trust level (to avoid excessively high trust levels for certain types of propositions).

[0100] The original process required splitting and accumulating the 12 composite subsets one by one, which was redundant and time-consuming, and did not meet the requirement of "millisecond-level feedback" for drilling exploration (the calculation should be completed within 20ms).

[0101] This invention simplifies the above process to improve computational efficiency. The core logic of the simplification is: in drilling exploration scenarios, the confidence level of composite subset propositions is inherently low (or negligible), so there is no need to split them one by one. The confidence level of the composite subset is directly allocated "in a fixed proportion" to its contained "highly relevant single subsets," or the extremely small composite confidence level is directly discarded. Specifically, it includes 3 steps:

[0102] 1. Filter for "meaningful composite subsets" and eliminate redundant propositions.

[0103] The detection target of this invention is a "definite geological body" (fault, fracture, etc.), and most composite subset propositions (such as { , The statement "either a fault or no anomaly" has no practical significance in engineering and its reliability is almost zero. Therefore, only two types of "highly relevant composite subsets" are retained: { , } (Macroscopic structure + microscopic anomaly), { (Microscopic anomalies + water-bearing fractures) – These two types of composite propositions have relatively high confidence levels, and the subsets they contain are the core objectives of concern in engineering; the other 10 composite subsets (such as { , }、{ The trust level of}) is directly regarded as "0", skipping the splitting step and greatly reducing the amount of computation.

[0104] 2. Simplified allocation rules: Allocation is based on a fixed ratio according to project risk.

[0105] In existing technologies, the "redistribution of confidence in composite subset propositions" in DS evidence theory adopts a "proportional splitting" approach. Its core characteristics are a lack of contextualization, numerous computational steps, and reliance on dynamic proportional adjustments, specifically manifested in:

[0106] (1) The proportions are not tied to any specific scenario: the proportions are mostly "average distribution" or "dynamic proportions based on historical data," without being combined with the core needs of the specific detection scenario. For example, the compound proposition { (Fault) The current technology for splitting the trust level of (fractures) may use a "5:5 average distribution" or needs to dynamically adjust the ratio by calculating the historical identification accuracy of the two types of geological bodies in real time (e.g., split according to a 6:4 ratio if the historical accuracy is 6:4). However, this ratio does not take into account the "geological risk priority in drilling exploration". If a fault is not identified, it will cause the overall borehole trajectory to deviate, and the risk is much higher than that of fractures. Blindly splitting according to the average or historical ratio may dilute the trust level of high-risk targets and affect engineering decisions.

[0107] (2) Redundant calculation steps: First, the number of all subset propositions included in the composite proposition / historical accuracy needs to be counted. Then, the splitting ratio of each subset is calculated using a formula. Finally, the confidence level is accumulated one by one. (Based on the identification framework) For example, only one composite proposition containing three subsets (such as...) To achieve this, the historical weights of three subsets must first be determined, and then the trust level must be split proportionally. Each step takes about 15ms to calculate, and the total time for 12 composite propositions exceeds 180ms, which is completely insufficient to meet the "millisecond-level feedback" requirement of drilling exploration.

[0108] (3) Lack of engineering orientation in results: Existing technology breakdown only pursues "mathematical probability equilibrium" and does not consider the "risk-first" logic of drilling operations. For example, the compound proposition { (Ordinary fissures) (Water-bearing fractures)}, existing technologies may divide them into "5:5 averages", but water-bearing fractures are prone to borehole collapse, and the risk is much higher than that of ordinary fractures. This division will underestimate the reliability of water-bearing fractures and may cause safety hazards.

[0109] The "simplified splitting rule" of this invention does not innovate the "proportional splitting" itself, but rather addresses the core needs of drilling exploration scenarios (risk priority, real-time feedback) by transforming "dynamically complex proportions" into "fixed priority proportions." By "binding fixed values ​​to scenarios," it reduces calculation steps while simultaneously enhancing the engineering value of the results. The specific simplification path is as follows:

[0110] For the two types of composite subsets that are retained, this invention does not rely on historical data or real-time statistics, but is based on the "risk-first" logic of drilling operations, and presets a fixed split ratio, completely eliminating the redundant step of "dynamic ratio calculation":

[0111] Compound propositions { , (Fault + Common Fracture): The geological risk of faults (causing borehole trajectory deviation) is much higher than that of common fractures. The confidence level should be allocated at a 6:4 ratio (e.g., m ({ , })=0.2, 0.12 is obtained. (Result: 0.08); Compound Proposition { (Ordinary fractures + water-bearing fractures): The risk (hole collapse) of water-bearing fractures is much higher than that of ordinary fractures. The confidence level should be allocated at a ratio of 3:7 (e.g., m ({ })=0.15, The result is 0.045. (Result: 0.105). Other fixed ratios can also be set according to actual needs.

[0112] The core advantage of this design is that the splitting rules are changed from "dynamic calculation" to "table lookup". The splitting time of a single compound proposition is reduced from 15ms to less than 1ms, and the results directly match the engineering risk requirements. High-risk targets gain higher trust and avoid the problem of existing technologies being "mathematically balanced but engineering-failed".

[0113] 3. Omit the "secondary verification" and directly summarize the output.

[0114] The original process performs a "consistency check" on the trust scores after splitting (to prevent certain types of propositions from having excessively high trust scores). This invention, through "hardware noise reduction + data weight optimization," has already ensured the reliability of the input data in advance. Therefore, after splitting, the trust scores are directly accumulated to the corresponding single subsets without verification; finally, only the trust scores of the four single subsets are output (e.g., m( )=0.62、m ( ) = 0.225, m ( ) = 0.105, m ( ()=0.05), which is directly used for Pignatic probability transformation to output the optimal geological conclusion.

[0115] The essence of this invention's simplification is not "trading speed for lower accuracy," but rather "customized trade-offs for drilling scenarios": discarding composite subsets without engineering significance reduces the number of computational objects; replacing complex proportional splitting with fixed-ratio allocation rules based on engineering risk reduces computational steps; and omitting redundant verification shortens the process. Using the simplified process, the computation cycle can be compressed to 18ms, significantly improving computational efficiency.

[0116] Step 7: Based on the weights of seismic data and radar data and Adjust the confidence level of each subset proposition: , , The confidence level corresponding to earthquake data. The level of trust in the radar data. , This is the revised trust level.

[0117] Step 8: Calculate the fusion confidence of all propositions using the DS combination rule based on the revised confidence level to determine the detection results.

[0118] Furthermore, it also includes step 9: calculating the confidence interval based on the fused confidence value to determine the reliability level of the fusion result.

[0119] In this embodiment, the fusion results are output in the form of "geological conclusion + confidence level". A confidence level of ≥85% is a highly reliable conclusion, which is directly used to guide drilling decisions; a confidence level of 60% ≤ confidence level < 85% is a moderately reliable conclusion, which is further verified in combination with drilling speed (such as reducing to 70% of the conventional speed); a confidence level of <60% is a low-reliability conclusion, which initiates secondary exploration and data collection.

[0120] Step 10: Make engineering decisions based on the fusion results and reliability levels.

[0121] Example: When a horizontal borehole of a kilometer-scale depth reaches a depth of 820m, seismic data... =0.71、 =0.15、 =0.08、 =0.06; Radar data =0.63、 =0.21、 =0.12、 =0.04; Data quality test result =32dB, P=94%, weighting coefficient =0.52、 =0.48; calculated by fusion =0.89、 =0.07、 =0.03、 =0.01, confidence interval [0.89, 1.0], judged as a highly reliable fault structure, the ground system immediately adjusted the borehole trajectory to avoid the area.

[0122] Conventional DS algorithms directly use raw data acquired during drilling, failing to consider issues such as decreased signal-to-noise ratio of seismic waves and polarization distortion of radar waves caused by drilling vibrations. Low-quality data easily dominates the fusion results. This invention, however, employs a "hardware denoising + algorithm weighting" mechanism to ensure the reliability of the evidence. Addressing fluctuations in the quality of drilling data, it first performs physical denoising and then dynamically adjusts the weights of seismic and radar data, ensuring the reliability of the fused input data. Conventional DS algorithms require the redistribution of confidence levels for a large number of composite subset propositions, resulting in cumbersome calculations (approximately 100ms cycle) that cannot meet the millisecond-level feedback requirements of drilling exploration. This invention, by deleting composite subset propositions with low confidence levels, omits "secondary verification," effectively improving processing efficiency.

[0123] Correspondingly, the present invention also provides an in-situ underground detection system suitable for kilometer-scale horizontal boreholes, comprising:

[0124] Data acquisition unit: used for acquiring seismic and radar data while drilling;

[0125] Data processing unit: Processes the data collected by the data acquisition unit based on the underground in-situ detection method applicable to kilometer-level horizontal boreholes.

[0126] Specifically, such as Figure 3 As shown, the underground in-situ exploration system suitable for kilometer-scale horizontal boreholes consists of four core components: a drilling subsystem, an equipment mounting subsystem, a power cable subsystem, and a three-dimensional collaborative exploration while drilling subsystem. The drilling subsystem is responsible for drilling kilometer-scale horizontal boreholes in the underground rock strata; the equipment mounting subsystem is used to install and fix the controllable mechanical source fine-grained seismic exploration system and the fully polarized borehole radar transceiver antenna, among other exploration equipment; the power cable subsystem provides a stable power supply and a reliable signal transmission channel for the entire platform and exploration equipment; and the three-dimensional collaborative exploration while drilling subsystem, through data fusion algorithms based on D-S evidence theory, achieves three-dimensional and fine-grained exploration of macroscopic geological structures and small geological anomalies. These four components work together to achieve integrated exploration while drilling functionality.

[0127] (1) Drilling subsystem

[0128] The drilling subsystem adopts a modular design, including the drill bit and drill pipe, power drive unit and guidance device.

[0129] Drill Bits and Drill Pipes: Drill bits are designed differently for various geological conditions. In hard rock formations, insert cone bits or diamond core bits are used. The carbide teeth of insert cone bits can effectively break hard rock, while diamond core bits can obtain complete core samples. In soft soil formations, PDC bits are used, whose polycrystalline diamond composite blades have excellent cutting performance. Drill pipes are made of high-strength alloy materials, specifically chromium-molybdenum alloy steel, and undergo a special heat treatment process to achieve high strength, high toughness, and fatigue resistance. A special threaded connection method, the double-shoulder thread connection, is used between drill pipes, which not only ensures connection strength but also improves sealing performance and prevents drilling fluid leakage.

[0130] Power drive unit: Utilizing a combination of a high-power hydraulic motor, a variable displacement pump, and a proportional valve, the torque and speed can be adjusted in real-time according to the formation hardness. The high-power hydraulic motor serves as the power source, with its rated power selected based on the drilling requirements of kilometer-level horizontal holes, providing sufficient torque and speed. The hydraulic system is equipped with a variable displacement pump and a proportional valve. By adjusting the displacement of the variable displacement pump and the opening of the proportional valve, the output power and speed of the hydraulic motor can be precisely controlled to adapt to drilling requirements under different geological conditions. For example, when encountering hard rock formations, the torque of the hydraulic motor is increased; in soft soil formations, the speed is appropriately increased to improve drilling efficiency.

[0131] Guiding System: Integrates Measurement While Drilling (MSD) and Logging While Drilling (LOD) technologies. The MSD system measures parameters such as borehole inclination, azimuth, and tool face angle in real time using sensors installed inside the drill pipe. The LOD system measures geological parameters such as formation resistivity and natural gamma ray, transmitting the data to the surface control system. Based on the preset borehole trajectory and measurement data, the surface control system adjusts the drill bit's drilling direction by controlling the steering and propulsion speed of the hydraulic motors, achieving precise control of the borehole trajectory and ensuring the borehole accurately reaches the predetermined position.

[0132] (2) Equipment-mounted subsystem

[0133] Structural Design: Utilizing a high-strength aluminum alloy frame and standardized quick-plug interfaces, the design balances lightweight construction with efficient equipment loading and unloading. An external nano-composite protective coating adapts to complex underground environments. The frame, constructed from high-strength aluminum alloy, is lightweight yet strong. Multiple standardized equipment mounting interfaces with quick-plug designs facilitate the installation and removal of detection equipment. Internally, the frame incorporates a vibration damping module, a core hardware support for ensuring the quality of drilling data. This module consists of a multi-stage damping structure and a vibration adaptive adjustment unit, designed to specifically counteract high-frequency vibrations (10-500Hz) and low-frequency impacts (≤10Hz) generated during kilometer-scale horizontal hole drilling, reducing vibration interference to seismic and radar detection equipment at its source and ensuring data reliability.

[0134] The overall structure of the shock absorption module is as follows:

[0135] ① Core damping layer

[0136] Primary vibration damping: Frame buffer layer. A vibration damping pad made of high-damping natural rubber and carbon fiber composite is laid between the equipment mounting frame and the detection equipment mounting base. The thickness is customized according to the equipment weight (5-12mm). This vibration damping pad has an elastic modulus of 1.2-1.8MPa and a damping ratio ≥0.35. It can directly absorb high-frequency vibrations generated during drilling (mainly from drill pipe rotation friction), with a vibration attenuation rate of ≥75% in the 100-500Hz frequency band. It is the basic buffer unit of the vibration damping system.

[0137] Secondary vibration damping: Equipment suspension layer. For critical equipment such as controllable mechanical seismic sources and seismic detectors, a suspension damping structure of "metal springs + dampers" is adopted. The metal springs are made of fatigue-resistant alloy material, and the stiffness coefficient can be adjusted according to the equipment weight (50-200 N / mm). Combined with hydraulic dampers, vibration energy is dissipated, achieving an attenuation rate of ≥85% for mid-frequency vibrations of 10-100Hz. The natural frequency of the seismic detector's dedicated suspension seat is ≤5Hz, avoiding resonance with drilling vibrations and ensuring accurate capture of weak seismic reflected waves.

[0138] ②Auxiliary damping unit

[0139] Vibration Isolation Cover: A flexible vibration isolation cover is installed outside the fully polarized borehole radar antenna array. Made of a composite material of silicone rubber and aramid fiber, it provides dual functions of waterproofing, corrosion resistance, and vibration isolation. The interior of the isolation cover is filled with porous vibration-absorbing foam, which further attenuates local vibrations caused by rock debris impacts and drilling fluid scouring within the borehole. It achieves an attenuation rate of ≥60% for high-frequency impacts, ensuring the polarization stability of the radar antenna.

[0140] Adaptive Adjustment Unit: This unit incorporates a miniature vibration sensor (response frequency 0.1-1000Hz, measurement range ±50g) and an electromagnetic adjustment module to monitor vibration intensity and frequency in real time during drilling. When the vibration sensor detects a vibration amplitude exceeding a preset threshold (≥0.8g), the electromagnetic adjustment module automatically adjusts the damping coefficient of the damper in the secondary damping structure (adjustment range 0.5-2.5Ns / mm), completing adaptive adaptation within 30ms. This ensures stable vibration reduction even under complex geological conditions (such as drilling in hard rock fracture zones).

[0141] Equipment Installation: The controllable mechanical seismic source fine-grained seismic detection system and the fully polarized borehole radar transceiver antenna are installed at different locations on the equipment mounting system. The controllable mechanical seismic source is fixed to the frame using a dedicated clamp with a hydraulic locking device to ensure the source remains stable during drilling. The fully polarized borehole radar transceiver antenna adopts an "array layout + adjustable angle design" to ensure stable positioning and uniform signal coverage during drilling. The array layout is installed on the side of the frame, and the transmission and reception of radar waves are optimized by adjusting the angle and spacing of the antennas. To protect the detection equipment from the corrosive effects of the complex underground environment, the equipment mounting system is wrapped with a waterproof, dustproof, and corrosion-resistant protective coating. The protective coating uses a nano-composite coating material, which has excellent sealing performance and anti-aging properties.

[0142] (3) Power cable conduit subsystem

[0143] Cables and signal transmission lines: The cables are made of highly flexible, wear-resistant special cables, capable of withstanding complex working conditions such as tension, bending, and torsion during kilometer-level horizontal hole drilling, providing a stable power supply for the drilling system and detection equipment. The signal transmission lines use a combination of optical fiber and coaxial cable. The optical fiber is used to transmit large amounts of detection data, offering advantages such as high transmission speed and strong anti-interference capability; the coaxial cable is used to transmit control signals, ensuring the stability and reliability of signal transmission.

[0144] Protective Sleeve and Retraction Device: The protective sleeve is made of high-strength metal-plastic composite tubing with internal spiral reinforcing ribs to improve its compressive strength and bending resistance. The protective sleeve encases the cable and signal transmission line, preventing mechanical damage and chemical corrosion during drilling. The power cable system also features an automatic retraction device, consisting of a motor, drum, and tension sensor. The motor automatically controls the drum's retraction speed based on the drilling depth and speed; the tension sensor monitors the tension of the cable and signal transmission line in real time, automatically adjusting the retraction speed when the tension exceeds a set threshold to prevent the cable from becoming too tight or too loose.

[0145] (4) Three-dimensional drilling collaborative exploration subsystem

[0146] Macroscopic Geological Structure Detection (Controllable Mechanical Source Drilling-While-Deepening Refined Seismic Detection System): This system includes a controllable mechanical source, a seismic detector, and a data acquisition and processing unit. The controllable mechanical source uses a hydraulically driven vibration device. By adjusting the pressure and flow rate of the hydraulic system, the vibration frequency, amplitude, and waveform of the source can be precisely controlled. During drilling, the controllable mechanical source generates seismic waves according to preset parameters. These waves propagate through the underground medium and are reflected and refracted when they encounter different geological interfaces. The seismic detector uses a high-sensitivity accelerometer, mounted on the equipment system, to receive the reflected and refracted seismic wave signals. The data acquisition and processing unit amplifies, filters, and digitizes the signals acquired by the seismic detector, and transmits the data to the ground control system via a signal transmission line. The ground control system uses advanced seismic migration imaging algorithms to process and analyze the acquired data, generating three-dimensional images of the underground geological structure, enabling refined detection of large geological structures such as faults and karst caves.

[0147] Detection of Small Geological Anomalies (Fully Polarized Borehole Radar Transceiver Antenna): The fully polarized borehole radar transceiver antenna consists of an antenna array composed of multiple horizontally polarized and vertically polarized antennas. The antenna employs microstrip antenna technology, featuring small size, light weight, and high radiation efficiency. During drilling, the radar transceiver antenna emits high-frequency electromagnetic waves into the surrounding strata. These electromagnetic waves are reflected and scattered when they encounter different geological bodies. The reflected and scattered electromagnetic waves are received by the antenna and transmitted to the ground-based data acquisition and processing unit via signal transmission lines. The data acquisition and processing unit analyzes the received radar signals, extracting information such as the amplitude, phase, and polarization of the radar waves. Using this information, combined with ground-penetrating radar inversion algorithms, it images the underground geological structure, clearly displaying the distribution of small geological anomalies such as fractures and aquifers around the borehole.

[0148] Collaborative Detection Mechanism: A controllable mechanical source precision seismic detection system and a fully polarized borehole radar transceiver antenna work collaboratively during drilling. The seismic detection system is primarily used to detect geological structures ahead of the borehole and over a large area, providing macroscopic geological guidance for drilling, such as faults and karst caves. The borehole radar system performs high-resolution detection of a smaller area around the borehole (usually within a range several times the borehole diameter), promptly identifying potential geological risks such as fissures and aquifers. The data collected by both systems are fused and processed in the ground control system. Through a data fusion algorithm based on D-S evidence theory, the advantages of seismic and radar data are complemented to form a more accurate and comprehensive subsurface geological information model, enabling three-dimensional and precise detection of subsurface geological structures.

[0149] This invention effectively improves processing efficiency by "simplifying the algorithm (compressing the calculation cycle to 18ms) + fiber optic transmission (delay ≤1ms)," meeting the requirement of "millisecond-level feedback guiding drilling" for drilling while drilling; and by "enhancing data dimensionality (3D → 6D) with fully polarimetric radar + expanding the recognition framework (adding new features)." The "Proposition" addresses the problem of "insufficient dimensionality of dual-source data (seismic + radar) leading to low discrimination" in drilling scenarios, improving the accuracy of identifying special geological targets (such as water-bearing fractures); by synchronizing drilling and exploration, the cycle is shortened to 15-20 days, increasing efficiency by more than 50%, while avoiding equipment wear and time waste caused by secondary well runs, thus solving the industry pain point of "slow project progress".

Claims

1. A method for in-situ underground detection of a horizontal hole of the order of a kilometre, characterised in that, Comprise: Step 1, using the data acquisition unit based on hardware noise reduction while drilling to collect seismic data and radar data; Step 2, acquiring characteristic parameters based on the collected seismic data and radar data, wherein the characteristic parameters acquired based on the seismic data at least include seismic wave signal-to-noise ratio , and the characteristic parameters acquired based on the radar data at least include radar wave polarization purity P; Step 3, determining the reliability of the data according to the signal-to-noise ratio of the seismic wave , radar wave polarization purity P and a preset reliability division rule. Step 4, determining weights of seismic data and radar data according to reliability of data and , and ; Step 5, defining the recognition framework and obtaining its power set, the power set including single subset propositions and compound subset propositions, wherein, is a macro-structure, is a micro-anomaly, is a vertically polarized reflection strong = water-filled fracture, is no significant geological anomaly; Step 6, the characteristic parameters obtained are matched with the characteristic propositions to obtain the trust degree of each proposition, and the composite subset propositions are deleted according to the trust degree; the trust degree of the remaining composite subset propositions is distributed to the corresponding single subset propositions by a fixed proportion according to the engineering risk; Step 7, modifying the trust degree of each single subset proposition according to the weight of seismic data and radar data and , , , is the trust degree corresponding to seismic data, is the trust degree corresponding to radar data, , is the modified trust degree; Step 8, based on the corrected trust degree, the fusion trust degree of all propositions is calculated using the D-S combination rule to determine the detection result.

2. The method for in-situ underground detection of a horizontal hole of a kilometer order of magnitude according to claim 1, characterized in that, The hardware noise reduction mode is to set shock-absorbing rubber pads, suspension damping structures based on metal springs + dampers and / or vibration isolation covers.

3. The method for in-situ underground detection of a horizontal hole of a kilometer order of magnitude according to claim 1, characterized in that, The preset reliability division rule is: high reliability, medium reliability, low reliability, a high reliability threshold corresponding to a seismic wave signal-to-noise ratio, a low reliability threshold corresponding to a seismic wave signal-to-noise ratio. high reliability, medium reliability, low reliability, high reliability threshold corresponding to radar wave polarization purity, low reliability threshold corresponding to radar wave polarization purity.

4. The method for in-situ underground detection of a horizontal hole of a kilometer order of magnitude according to claim 3, characterized in that, If and then reacquire seismic data and radar data; If and then , ; If and then , ; Otherwise, first according to the signal-to-noise ratio of seismic wave Determination of seismic data and radar data with radar wave polarization purity P the base weight of the basis weight and ; then the base weight of the basis weight and is calculated and : , .

5. The method for in-situ underground detection of a horizontal hole of a kilometer order of magnitude according to claim 4, characterized in that, base weight and Specifically: When and , , ; When and , , ; When and , , ; When and , , .

6. The method for in-situ underground detection of a horizontal hole of a kilometer order of magnitude according to claim 3, characterized in that, The high reliability threshold value corresponding to the seismic wave signal-to-noise ratio is 30dB, and the low reliability threshold value is 15dB; the high reliability threshold value corresponding to the radar wave polarization purity is 90%, and the low reliability threshold value is 80%.

7. The method for in-situ underground detection suitable for kilometer- scale horizontal boreholes according to claim 1, characterized in that, The composite subset proposition retains only }, }.

8. The method for in-situ underground detection of a horizontal hole of a kilometer scale according to claim 1, characterized in that, Further comprising: step 9, calculating the confidence interval based on the fused trust degree value to determine the reliability level of the fusion result.

9. The method for in-situ underground detection of horizontal holes of km scale according to claim 8, characterized in that, Further comprising: Step 10, according to the fusion result and the reliability level, making engineering decisions.

10. An underground in-situ probing system suitable for kilometer- scale horizontal boreholes, characterized in that, Comprise: Data acquisition unit: used for collecting seismic data and radar data while drilling; Data processing unit: processing the data collected by the data acquisition unit based on the underground in-situ detection method suitable for kilometer-level horizontal holes according to any one of claims 1-9.

Images