A method and system for real-time identification of surrounding rock structural plane while drilling based on peeping calibration

Abstract

A surrounding rock structure face real-time identification method and system based on peeping calibration while drilling, the method: load the structure face identification operation package while drilling, only real-time acquisition of drilling parameters in the drilling process; update the drilling parameter benchmark response online, calculate the normalized mechanical response characteristics, and adaptively update the structure face abnormal trigger threshold; when the trigger condition is met, start the boundary fine positioning window, output the identification probability; determine the structure face type, start and end depth, development length and identification confidence through continuous window probability fusion, and mark the low confidence interval for peeping review of the drilled hole. The system: based on the real-time acquisition module of drilling parameters, the streaming window management module, the stable section benchmark update module, the mechanical response characteristic calculation module, the adaptive threshold update module, the boundary probability reasoning module, the multi-window fusion module, the confidence and review marking module, and the incremental sample management module form the identification system. The present application can improve the drilling peeping calibration accuracy and real-time identification efficiency while drilling.

Description

Technical Field

[0001] This invention belongs to the field of underground engineering surrounding rock detection, drilling measurement data processing, borehole inspection calibration and intelligent identification technology, specifically involving a method and system for real-time identification of surrounding rock structural surfaces based on inspection calibration during drilling. Background Technology

[0002] During the construction of underground engineering projects such as coal mine roadways and tunnels, various structural surfaces, including fissures, delamination, and rock strata interfaces, are commonly found within the surrounding rock. The spatial distribution and extent of these structural surfaces directly determine the integrity and bearing capacity of the surrounding rock, serving as the core basis for evaluating the stability of the roadway surrounding rock and optimizing the design of support parameters. Therefore, the rapid and accurate identification of the types and extent of structural surfaces in the surrounding rock is of significant practical engineering importance for ensuring safe construction of underground engineering projects and mitigating the risk of surrounding rock instability.

[0003] Currently, the identification of surrounding rock structural surfaces in underground engineering mainly relies on two methods: borehole inspection and drilling-while-drilling (DWD) technology. Both technologies have their advantages and disadvantages, but both suffer from the limitation of being unable to simultaneously achieve accurate calibration and real-time identification. Borehole inspection technology can directly obtain the morphology of borehole wall structural surfaces, fracture boundaries, rock strata interfaces, and the development range of fracture zones, providing reliable and accurate imaging results that can serve as a precise basis for calibrating the true location and type of surrounding rock structural surfaces. However, this technology must be implemented after drilling is completed, making it unsuitable for real-time monitoring during the drilling process. Furthermore, the detection results are easily affected by factors such as image quality, borehole mud obstruction, borehole wall cleanliness, and human interpretation experience, making it difficult to guarantee identification accuracy and stability, and thus preventing dynamic real-time identification throughout the entire drilling process.

[0004] Detection while drilling (DWP) technology can collect multi-dimensional drilling parameters such as drilling depth, drilling speed, rotational speed, thrust, and torque in real time during drilling, enabling continuous and dynamic characterization of the interaction between the drill bit and the surrounding rock. When the drill bit crosses a structural surface in the surrounding rock, various DWP parameters will exhibit significant abrupt changes or abnormal fluctuations, effectively supporting real-time dynamic identification of the structural surface. However, relying solely on a single DWP parameter for identification is highly susceptible to factors such as fluctuations in on-site drilling conditions, drill bit wear, differences in human operation, sensor noise, and local lithological changes. This can lead to frequent problems such as parameter anomalies, false triggering, structural surface boundary positioning misalignment, and poor stability of identification results, and it is also impossible to effectively evaluate the confidence level of the identification results.

[0005] In existing technologies for identifying surrounding rock structural surfaces, borehole inspection and real-time identification while drilling are largely independent processes, lacking an efficient integrated mechanism. Borehole inspection is used only for post-drilling imaging and manual result verification, while real-time identification relies solely on parameter anomalies for preliminary judgment. The industry has yet to develop an integrated operational solution that uses borehole inspection data as a priori calibration basis and relies on real-time identification based on drilling parameters during the drilling phase. Specifically, existing technologies lack several core adaptation mechanisms, including an online update mechanism for stable drilling sections, an adaptive anomaly threshold triggering mechanism adapted to field conditions, a two-level processing mechanism for candidate coarse positioning windows and precise boundary positioning windows for structural surfaces, and supporting technologies for multi-window identification probability fusion and confidence verification marking of identification results. This results in poor adaptability to operating conditions, low positioning accuracy, and insufficient reliability of identification results, making it difficult to meet the high-precision, real-time, and intelligent identification requirements of surrounding rock structural surfaces in complex underground engineering projects. Therefore, there is an urgent need to provide a real-time identification method and system for surrounding rock structural surfaces based on borehole inspection calibration while drilling. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention provides a method and system for real-time identification of surrounding rock structural surfaces during drilling based on borehole inspection calibration. This method balances the accuracy of borehole inspection calibration with the efficiency of real-time identification during drilling, enabling a complete transformation from high-precision borehole inspection calibration to efficient real-time identification of drilling parameters. This provides a reliable technical solution for rapid on-site detection and intelligent identification of surrounding rock structural surfaces in underground engineering. Based on a modular collaborative architecture, this system effectively solves the problems in existing technologies, such as the difficulty of real-time application of borehole inspection, the significant impact of drilling condition fluctuations on independent identification during drilling, insufficient probabilistic output of structural surface boundaries, and the lack of a verification and marking mechanism for low-confidence identification results.

[0007] To achieve the above objectives, the present invention provides a method for real-time identification of surrounding rock structural surfaces during drilling based on observation calibration, comprising the following steps: Step 1: Obtain the drilling structural surface identification operation package pre-calibrated using borehole inspection data; calibrate the drilling structural surface identification model using borehole inspection data during the model training or calibration phase to form an operation package that relies solely on drilling parameters during the field identification phase. Step 2: Real-time acquisition of drilling parameters and construction of a flow-based drilling window; During the drilling process on site, real-time acquisition of drilling parameters, construction of a flow-based drilling window according to the preset depth step size, and determination of stable drilling sections from the drilled sections; Step 3: Update the baseline response of the drilling parameters online and calculate the normalized mechanical response characteristics; based on the stable drilling section in the current drilled section of the borehole, update the baseline response of the drilling parameters online and calculate the normalized mechanical response characteristics within the flow drilling window; Step 4: Adaptively update the structural surface anomaly trigger threshold; dynamically update the structural surface anomaly trigger threshold for each normalized mechanical response feature based on the mechanical response characteristic statistics of the current stable drilling section of the borehole. Step 5: Anomaly Triggering and Boundary Fine Positioning Window Activation; When the normalized mechanical response characteristics meet the anomaly triggering conditions, the flowing drilling window is marked as a candidate triggering window, and the boundary fine positioning window centered on the candidate triggering window is activated; Step 6: Model Inference and Multi-Probability Output; Input the drilling parameter sequence and normalized mechanical response features within the boundary fine-tuning window into the drilling structure surface identification model, and output the structure surface type probability, the starting boundary probability distribution, the ending boundary probability distribution, and the structure surface existence probability. Step 7: Multi-window probabilistic fusion and structural surface parameter determination; perform probabilistic fusion on the output results of multiple consecutive boundary fine-localization windows to determine the structural surface type, starting depth, ending depth and development length of the same structural surface candidate interval; Step 8: Calculate the confidence level and mark the area to be reviewed; calculate the confidence level based on multi-dimensional information and output the structural surface recognition result; when the confidence level is lower than the preset review threshold, mark the corresponding depth interval as the area to be drilled for inspection and review.

[0008] As a preferred embodiment, the process of obtaining the pre-calibrated drilling structure surface identification runtime package in step 1 is as follows: During the model training or calibration phase, borehole inspection data is used to determine the structure type, starting depth, and ending depth of the structure. Combined with the drilling parameter sequence and mechanical response characteristics within the corresponding depth range, a drilling-only input model calibrated using borehole inspection data is trained and used as the drilling structure identification model. This model uses only the drilling parameter sequence and normalized mechanical response characteristics as input during the field identification phase. After training or calibration, the model parameters, standardized parameters for drilling parameter characteristics, stable drilling section determination parameters, initial threshold for abnormal triggering, boundary fine-positioning window parameters, multi-window fusion parameters, and confidence calculation parameters are encapsulated into a drilling structure identification runtime package. The drilling structure identification runtime package includes at least the drilling structure identification model parameters, drilling parameter feature standardization parameters, stable drilling section determination parameters, abnormal trigger initial threshold, boundary fine positioning window parameters, multi-window fusion parameters, and confidence calculation parameters; wherein, the borehole inspection data is only used in the training, calibration, or verification phases to determine the structure type, structure start depth, and structure end depth.

[0009] As a preferred option, in step 2, the process of acquiring drilling parameters in real time and constructing a streaming drilling window is as follows: S21: During the on-site drilling process, the drilling parameters of the borehole to be identified are collected in real time; the drilling parameters include multiple parameters such as drilling depth, drilling time, drilling speed, rotation speed, thrust, torque, drill pipe displacement, hydraulic pressure, drilling vibration, or drilling rig power; the on-site identification stage only uses the drilling parameters collected in real time during the drilling process to complete the structural surface identification; S22: Load the drilling structure surface identification and operation package into the field operation host, edge computing terminal or drilling rig data processing unit; divide the drilling parameters into continuous flow drilling windows according to the preset depth step size, and each flow drilling window corresponds to a continuous drilling depth range; S23: In the currently drilled section, if the mechanical response fluctuation amplitude in a certain depth section is lower than the preset stability threshold and is not marked as a candidate section for structural surface, then it is determined as a stable drilling section; the mechanical response fluctuation amplitude is determined based on the coefficient of variation or standard deviation of at least one of the parameters of drilling speed, rotation speed, thrust and torque within the depth window.

[0010] As a preferred embodiment, in step 3, the process of updating the online drilling parameter baseline response and calculating the normalized mechanical response characteristics is as follows: S31: Based on the stable drilling section in the current drilled section, the baseline response of the drilling parameters is updated online; the baseline response of the drilling parameters is determined according to the mean, median, standard deviation, quantile or slip statistic of the drilling parameters in the stable drilling section; as the drilling progresses and new stable drilling sections are identified, the baseline response of the drilling parameters is continuously updated online to adapt to the gradual change of the surrounding rock mechanical properties in different depth sections; S32: Calculate the normalized mechanical response characteristics within the flow-while-drilling window; the normalized mechanical response characteristics include multiple of the following: energy per unit footage, mechanical specific energy, torque-to-thrust ratio, drilling rate residual, and energy mutation coefficient; wherein, energy per unit footage is the ratio of drilling energy to the corresponding footage; mechanical specific energy is the energy required to break a unit volume of rock; torque-to-thrust ratio is the ratio of torque to thrust; drilling rate residual is determined based on the difference between the measured drilling rate and the reference drilling rate predicted from the current thrust, rotational speed, and torque; and energy mutation coefficient is determined based on the normalized deviation of the energy per unit footage within the current flow-while-drilling window from the reference value of energy per unit footage in the stable drilling section. Among them, each of the above-mentioned normalized mechanical response characteristics is based on the reference response of the drilling parameters in the stable drilling section, and each mechanical response characteristic is converted to a dimensionless relative value based on the stable drilling section.

[0011] As a preferred embodiment, the adaptive update of the structural surface anomaly triggering threshold in step 4 is as follows: One or more of the following are used as mechanical response characteristic statistics: mean, standard deviation, median, quantile, moving mean, or moving median; within each flow-while-drilling window, the structural surface anomaly trigger threshold is dynamically updated based on the mechanical response characteristic statistics of the current stable drilling section of the borehole, as shown in the following formula; ; In the formula, For the first Structural anomaly triggering threshold of normalized mechanical response characteristics , For the current stable drilling section The median or mean of the normalized mechanical response characteristics. For the corresponding standard deviation, interquartile range, or absolute median, This is the threshold amplification factor.

[0012] As a preferred embodiment, the process of exception triggering and boundary fine positioning window activation in step 5 is as follows: S51: When the normalized mechanical response characteristics within the flow-while-drilling window meet the abnormal triggering conditions, the flow-while-drilling window is marked as a candidate triggering window. The abnormality triggering conditions include: [the following conditions are met] for a continuous period of not less than [number missing]. Within a given depth sampling point, at least one of the following: energy mutation coefficient, drilling rate residual, drilling stiffness change rate, peak mechanical energy, or torque-to-thrust ratio change... The characteristic exceeds the corresponding structural surface anomaly trigger threshold; S52: Activate the boundary fine positioning window centered on the candidate trigger window; the boundary fine positioning window is used to output the probability distribution of the starting boundary and the probability distribution of the ending boundary of the structural surface; the depth length of the boundary fine positioning window is greater than the depth length of the candidate trigger window, and its extension range extends to the front and rear sides of the candidate trigger window by a preset depth distance to ensure complete coverage of the boundary transition area of ​​the structural surface.

[0013] As a preferred embodiment, in step 6, the structure surface type probability is used to characterize the probability that the current boundary fine positioning window belongs to different structure surface types, and the sum of the probabilities of each type is 1; the starting point boundary probability distribution is used to characterize the probability that the starting depth of the structure surface is at each depth sampling point within the boundary fine positioning window, and the depth position corresponding to the probability peak is the most likely starting depth; the ending point boundary probability distribution is used to characterize the probability that the ending depth of the structure surface is at each depth sampling point within the boundary fine positioning window, and the depth position corresponding to the probability peak is the most likely ending depth; the structure surface existence probability is used to characterize the probability that a structure surface exists within the current boundary fine positioning window; each output is a dimensionless probability value, with a value range of [0,1].

[0014] As a preferred embodiment, the process of multi-window probability fusion and structural surface parameter determination in step 7 is as follows: S71: When the structural surface type output by adjacent boundary fine positioning windows has the same maximum probability category, and the probability distribution of the starting boundary or the probability distribution of the ending boundary overlaps in the depth direction, the adjacent boundary fine positioning windows are merged into the same structural surface candidate interval. The overlap determination condition is: the probability distribution of the end boundary of the previous window and the probability distribution of the start boundary of the next window intersect in the depth coordinate, and the probability values ​​of both in the overlapping area are greater than the preset overlap probability threshold. S72: Weighted fusion of the probability distributions of the starting boundaries of multiple fine-positioning windows within the same candidate interval of the structural surface, and determination of the starting depth of the structural surface by the peak probability or expected probability position after fusion; weighted fusion of the probability distributions of the ending boundaries, and determination of the ending depth of the structural surface by the peak probability or expected probability position after fusion; the development length of the structural surface is obtained by subtracting the starting depth of the structural surface from the ending depth of the structural surface. The weights of the weighted fusion are determined based on one or more of the following: the probability of the structural surface type of the boundary fine-positioning window, the intensity of the physical response anomaly, the distance between the center depth of the window and the center depth of the candidate interval, and the completeness rate of the drilling parameters within the window. The larger the weight value, the more reliable the identification result of the window. The intensity of the physical response anomaly is the sum or peak value of the normalized mechanical response features within the boundary fine-positioning window that exceed the corresponding trigger threshold.

[0015] As a preferred embodiment, the process of identifying the confidence level calculation and the marker to be reviewed in step 8 is as follows: S81: Calculate the identification confidence level based on the probability of structural surface type, the peak probability of the starting boundary, the peak probability of the ending boundary, the consistency of the continuous window, and the intensity of physical response anomalies. ,in The base confidence level is determined by the probability of the structural surface type, the peak probability of the starting boundary, and the peak probability of the ending boundary. This is the window stability correction factor determined by the consistency of continuous windows; This is a physical support correction factor determined by the intensity of the physical response anomaly. S82: When the identification confidence level is lower than the preset verification threshold, or the peak width of the probability distribution of the starting point boundary and the probability distribution of the ending point boundary is greater than the preset uncertainty threshold, the corresponding depth interval is marked as the borehole inspection verification interval; the peak width is the depth interval width in the probability distribution where the probability value is greater than the preset probability ratio threshold; S83: When the deviation between the subsequent borehole inspection verification result and the structure surface identification result in terms of structure surface type, structure surface start depth, or structure surface end depth exceeds a preset deviation threshold, the drilling parameter sequence and verification result within the corresponding depth interval are added to the incremental sample library for subsequent incremental training or parameter updates of the model.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention pre-calibrates a drilling-while-drilling (DWDD) structural surface identification runtime package using borehole inspection. This package includes complete identification configurations such as model parameters, normalization parameters, trigger thresholds, and fusion parameters. This allows for real-time structural surface identification based on DWDD parameters without requiring real-time input of borehole inspection images, videos, or borehole wall unfolding diagrams during the field phase. The training phase fully utilizes the accurate calibration capabilities of borehole inspection for structural surface types and boundary locations, while the inference phase relies solely on DWDD parameters for identification. This achieves a balance between calibration accuracy and identification efficiency, significantly improving continuous field identification efficiency and engineering adaptability, while simultaneously reducing field hardware deployment costs and operational complexity.

[0017] 2. This invention, through online updating of the stable drilling section benchmark and adaptive anomaly triggering threshold, can adapt to changes in the drilling response benchmark under different boreholes, lithologies, and drilling conditions. The drilling parameter benchmark response is continuously updated online as drilling progresses and new stable drilling sections are identified, eliminating the influence of gradual changes in the mechanical properties of the surrounding rock at different depths on the mechanical response characteristics. The structural surface anomaly triggering threshold is dynamically updated based on the statistics of the current stable drilling section, adaptively determined by adding the threshold amplification factor to the mean or median and multiplying it by the standard deviation or interquartile range, so that the triggering conditions match the current drilling conditions, effectively reducing false alarms and missed alarms caused by the fixed threshold method under different geological and drilling conditions.

[0018] 3. This invention decouples structural anomaly detection from boundary probabilistic localization through a two-stage processing mechanism: a candidate trigger window and a fine-grained boundary localization window. The candidate trigger window performs coarse structural anomaly detection with fewer features and lower computational cost. The fine-grained boundary localization window is activated after extending a preset depth distance before and after the candidate window, outputting the probability distributions of the starting and ending boundary, thus achieving complete coverage of the transition region of the structural surface boundary. This two-stage mechanism significantly improves the accuracy of structural surface start and end depth localization while ensuring real-time performance.

[0019] 4. This invention, through multi-window probability fusion and recognition confidence calculation, can output structural surface type, starting depth, ending depth, development length, and recognition confidence. Multi-window fusion uses structural surface type probability, physical response anomaly intensity, and window location distance as weights to weight and fuse the boundary probability distributions of adjacent windows, effectively avoiding the influence of local noise interference on single-window recognition results. Recognition confidence integrates the base confidence, continuous window consistency correction factor, and physical response anomaly intensity correction factor, and verifies and marks low-confidence intervals, providing reliable data support for surrounding rock logging, support parameter optimization, and construction safety decisions. Furthermore, the deviation between the verification results and the recognition results triggers incremental sample library updates, forming a continuous improvement closed loop of recognition-evaluation-optimization.

[0020] 5. This invention uses an online update mechanism for the baseline response of drilling parameters to normalize the mechanical response characteristics with the stable drilling section as the reference benchmark. It converts absolute values ​​into dimensionless relative values, eliminating the influence of different drilling conditions and drilling depths on the absolute values ​​of mechanical response characteristics. This makes the normalized mechanical response characteristics have a unified comparison benchmark across the entire drilling depth range, significantly improving the consistency and reliability of cross-condition triggering of structural surface anomalies.

[0021] This method, through a technical approach including borehole inspection calibration package generation, online updating of stable drilling section benchmarks, adaptive threshold triggering, two-level window positioning, multi-window probability fusion, confidence assessment, and verification marking, balances borehole inspection calibration accuracy with real-time identification efficiency during drilling. It achieves a complete transformation from high-precision borehole inspection calibration to real-time and efficient identification of parameters during drilling, providing a reliable technical solution for rapid on-site detection and intelligent identification of surrounding rock structures in underground engineering. It is suitable for rapid detection of surrounding rock structures in coal mine roadways, tunnels, and underground engineering, and can be used for real-time on-site identification and verification marking of surrounding rock structures such as fissures, delamination, rock strata interfaces, and fracture zones.

[0022] This invention also provides a real-time identification system for surrounding rock structural surfaces based on peep calibration during drilling, used to implement a method for real-time identification of surrounding rock structural surfaces based on peep calibration during drilling, comprising: The calibration operation package calling module is used to call the drilling structure surface identification operation package that has been pre-calibrated using borehole inspection data; The real-time acquisition module for drilling parameters is used to acquire the drilling parameters of the borehole to be identified in real time during the drilling process on site. The streaming window management module is used to divide the real-time acquired drilling parameters into continuous streaming drilling windows according to a preset depth step size. Each streaming drilling window corresponds to a continuous drilling depth range. The stable section benchmark update module is used to update the benchmark response of drilling parameters online based on the stable drilling section in the current drilled section of the borehole. The mechanical response characteristic calculation module is used to calculate the normalized mechanical response characteristics within the flow-while-drilling window; The adaptive threshold update module is used to dynamically update the structural surface anomaly trigger threshold within each flow-while-drilling window based on the statistics of the current stable drilling section of the borehole, using the mean, standard deviation, median, quantile, moving mean, or moving median as the mechanical response characteristic statistics. The boundary probability inference module is used to input the drilling parameter sequence and normalized mechanical response features within the boundary fine positioning window into the loaded drilling structure surface identification model, and output the structure surface type probability, the starting boundary probability distribution, the ending boundary probability distribution, and the structure surface existence probability. The multi-window fusion module is used to probabilistically fuse the output results of multiple consecutive boundary fine-positioning windows; The confidence and verification labeling module is used to calculate and identify confidence based on the probability of structural surface type, the peak probability of the starting boundary, the peak probability of the ending boundary, the consistency of the continuous window, and the intensity of physical response anomalies. When the identification confidence level is lower than the preset verification threshold, or the peak width of the boundary probability distribution is greater than the preset uncertainty threshold, the corresponding depth interval is marked as the borehole inspection verification interval. The incremental sample management module is used to add the drilling parameter sequence and verification results within the corresponding depth range to the incremental sample library when the deviation between the structure type, starting depth or ending depth between the subsequent borehole inspection verification results and the structure identification results exceeds a preset deviation threshold.

[0023] In this invention, the configuration of the calibration operation package invocation module eliminates the need for real-time input of borehole peep images, videos, or borehole wall unfolding diagrams during the on-site identification phase, ensuring data independence and deployment convenience for real-time on-site identification from the source. The configuration of the real-time acquisition module for drilling parameters allows the use of drilling parameters acquired in real-time during the drilling process as the data source, eliminating the reliance on borehole peep data as real-time input. This frees on-site identification operations from dependence on borehole peep equipment, significantly improving identification efficiency during continuous drilling. The configuration of the streaming window management module automatically marks candidate trigger windows and initiates fine-grained boundary positioning windows. This two-level window mechanism of coarse detection and fine positioning ensures real-time performance while providing complete structural boundary coverage for boundary probabilistic inference. The configuration of the stable section benchmark update module effectively eliminates the influence of gradual changes in the mechanical properties of surrounding rock at different depths on mechanical response characteristics, providing a dynamic and reliable reference benchmark for the calculation of normalized mechanical response characteristics. The mechanical response feature calculation module eliminates the influence of different drilling conditions and borehole depths on the absolute values ​​of features, providing a unified comparison benchmark for the normalized mechanical response features across the entire borehole depth range. The adaptive threshold update module ensures that the trigger threshold accurately matches the actual working conditions and surrounding rock mechanical properties of the current borehole as drilling progresses, effectively avoiding false alarms and missed alarms under different geological conditions with fixed thresholds. The boundary probability reasoning module elevates structural surface identification from discrete point location to continuous probability representation, providing comprehensive probabilistic output for subsequent multi-window fusion and confidence quantification. The multi-window fusion module effectively overcomes the problem of local noise interference in single-window identification results, significantly improving the accuracy and robustness of structural surface boundary location. The confidence and verification marking module enables multi-dimensional quantitative evaluation of the reliability of structural surface identification results, and automatically marks the interval to be verified when the confidence level is below the threshold or the boundary uncertainty is too high, providing clear guidance for verification priorities in field operations. Based on the incremental sample management module, a continuous improvement process of identification, evaluation, verification, and optimization is formed, enabling the system to continuously learn and evolve from on-site feedback.

[0024] Based on a modular collaborative architecture, this system realizes the complete transformation from high-precision calibration of surrounding rock structural surfaces through borehole inspection to real-time and efficient identification of parameters while drilling. It can effectively solve the problems in existing technologies, such as the difficulty of real-time application of borehole inspection, the significant impact of working condition fluctuations on independent identification while drilling, insufficient probabilistic output of structural surface boundaries, and the lack of verification and marking mechanisms for low-confidence identification results. It provides a reliable, efficient, and evolvable system-level solution for rapid on-site detection and intelligent identification of surrounding rock structural surfaces in underground engineering. Attached Figure Description

[0025] Figure 1 This is a flowchart of the identification method in this invention; Figure 2 This is a flowchart of the abnormal triggering and boundary fine positioning window startup in this invention; Figure 3 This is a flowchart illustrating the confidence calculation and verification markers in this invention; Figure 4 This is a block diagram illustrating the principle of the identification system in this invention. Detailed Implementation

[0026] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0027] Example 1:

[0028] like Figure 1 As shown, this embodiment provides a method for real-time identification of surrounding rock structural surfaces during drilling based on observation calibration, including the following steps: Step 1: Obtain the drilling structural surface identification operation package pre-calibrated using borehole inspection data; calibrate the drilling structural surface identification model using borehole inspection data during the model training or calibration phase to form an operation package that relies solely on drilling parameters during the field identification phase. As a preferred option, the process of obtaining the drilling structure surface identification runtime package pre-calibrated from borehole inspection data is as follows: like Figure 1 As shown, during the model training or calibration phase, the type, starting depth, and ending depth of the structural surface are determined using borehole inspection data. Combined with the drilling parameter sequence and mechanical response characteristics within the corresponding depth range, a drilling-only input model calibrated using borehole inspection data is trained and used as the drilling structural surface recognition model. During the on-site recognition phase, this model uses only the drilling parameter sequence and normalized mechanical response characteristics as input, without using borehole inspection images, borehole videos, or borehole unfolding diagrams as real-time input. After training or calibration, the model parameters, standardized parameters for drilling parameter characteristics, stable drilling section determination parameters, initial threshold for abnormal triggering, boundary fine-positioning window parameters, multi-window fusion parameters, and confidence calculation parameters are encapsulated into a drilling structural surface recognition runtime package. The drilling structure identification runtime package includes at least the drilling structure identification model parameters, drilling parameter feature standardization parameters, stable drilling section determination parameters, abnormal trigger initial threshold, boundary fine positioning window parameters, multi-window fusion parameters, and confidence calculation parameters; wherein, the borehole inspection data is only used in the training, calibration, or verification phases to determine the structure type, structure start depth, and structure end depth. In this technical solution, the type, starting depth, and ending depth of the structural surface are accurately determined using borehole inspection data during the model training or calibration phase. This is combined with the drilling parameter sequence and mechanical response characteristics for the corresponding depth range to train a drilling-only input model calibrated using borehole inspection data. This transforms the high-precision calibration capability of borehole inspection for structural surfaces into parameterized knowledge for the drilling-while-identifying model. After training or calibration, the model parameters, feature standardization parameters, stable drilling section determination parameters, anomaly trigger initial threshold, boundary fine-positioning window parameters, multi-window fusion parameters, and confidence calculation parameters are uniformly encapsulated into a drilling-while-identifying structural surface runtime package. This allows the on-site identification phase to complete structural surface identification based on real-time drilling parameters simply by loading this runtime package, eliminating the need for real-time input of borehole inspection images, videos, or borehole wall unfolding diagrams. This runtime package encapsulation mechanism separates the high-precision calibration of borehole inspection during the training phase from the efficient identification of drilling parameters during the inference phase, balancing calibration accuracy with real-time on-site identification efficiency.

[0029] Step 2: Real-time acquisition of drilling parameters and construction of a flow-based drilling window; During the drilling process on site, real-time acquisition of drilling parameters, construction of a flow-based drilling window according to the preset depth step size, and determination of stable drilling sections from the drilled sections; As a preferred method, the process of acquiring drilling parameters in real time and constructing a streaming drilling window is as follows: S21: During the on-site drilling process, the drilling parameters of the borehole to be identified are collected in real time; the drilling parameters include multiple parameters such as drilling depth, drilling time, drilling speed, rotation speed, thrust, torque, drill pipe displacement, hydraulic pressure, drilling vibration, or drilling rig power; the on-site identification stage only uses the drilling parameters collected in real time during the drilling process to complete the structural surface identification; S22: In the on-site identification phase, the drilling-while-drilling structural surface identification package is loaded into the on-site operation host, edge computing terminal, or drilling rig data processing unit; wherein, in the on-site phase, borehole inspection data is not input in real time, but structural surface identification is completed only using drilling parameters acquired in real time during drilling; such as Figure 2 As shown, the drilling parameters are divided into continuous flow drilling windows according to the preset depth step size, and each flow drilling window corresponds to a continuous drilling depth range. S23: In the currently drilled section, if the mechanical response fluctuation amplitude in a certain depth section is lower than the preset stability threshold and is not marked as a candidate section for structural surface, then it is determined as a stable drilling section; the mechanical response fluctuation amplitude is determined based on the coefficient of variation or standard deviation of at least one of the parameters of drilling speed, rotation speed, thrust and torque within the depth window; In this technical solution, multi-dimensional drilling parameters such as drilling speed, rotational speed, thrust, and torque are collected in real time during the drilling process. A continuous-flow drilling window is constructed using a preset depth step size, achieving structured real-time organization of drilling data. After loading the drilling structure identification package, which is pre-calibrated using borehole inspection data, onto the field operator or edge computing terminal, the field identification phase does not require real-time input of borehole inspection data. Structure identification can be completed using only the drilling parameters collected in real time during drilling, achieving separation between high-precision borehole inspection calibration in the calibration phase and efficient drilling parameter identification in the field identification phase. Simultaneously, by identifying stable drilling sections in the currently drilled section where the mechanical response fluctuation amplitude is lower than a preset stability threshold, a dynamic and reliable reference benchmark is provided for the online updating of subsequent drilling parameter baseline responses and the calculation of normalized mechanical response characteristics, effectively adapting to changes in different drilling conditions and surrounding rock conditions.

[0030] Step 3: Update the baseline response of the drilling parameters online and calculate the normalized mechanical response characteristics; based on the stable drilling section in the current drilled section of the borehole, update the baseline response of the drilling parameters online and calculate the normalized mechanical response characteristics within the flow drilling window; As a preferred method, the process of updating the baseline response of drilling parameters online and calculating the normalized mechanical response characteristics is as follows: S31: Based on the stable drilling section in the current drilled section, the baseline response of the drilling parameters is updated online; the baseline response of the drilling parameters is determined according to the mean, median, standard deviation, quantile or slip statistic of the drilling parameters in the stable drilling section; as the drilling progresses and new stable drilling sections are identified, the baseline response of the drilling parameters is continuously updated online to adapt to the gradual change of the surrounding rock mechanical properties in different depth sections; S32: Calculate the normalized mechanical response characteristics within the flow-while-drilling window; the normalized mechanical response characteristics include multiple of the following: energy per unit footage, mechanical specific energy, torque-to-thrust ratio, drilling rate residual, and energy mutation coefficient; wherein, energy per unit footage is the ratio of drilling energy to the corresponding footage; mechanical specific energy is the energy required to break a unit volume of rock; torque-to-thrust ratio is the ratio of torque to thrust; drilling rate residual is determined based on the difference between the measured drilling rate and the reference drilling rate predicted from the current thrust, rotational speed, and torque; and energy mutation coefficient is determined based on the normalized deviation of the energy per unit footage within the current flow-while-drilling window from the reference value of energy per unit footage in the stable drilling section. Among them, each of the above-mentioned normalized mechanical response characteristics takes the reference response of the drilling parameters in the stable drilling section as the reference benchmark, and converts each mechanical response characteristic to a dimensionless relative value based on the stable drilling section, so as to eliminate the influence of different drilling conditions and drilling depth on the absolute value of the mechanical response characteristics. Step 4: Adaptively update the structural surface anomaly trigger threshold; Based on the mechanical response characteristic statistics of the current stable drilling section of the borehole, dynamically update the structural surface anomaly trigger threshold of each normalized mechanical response characteristic to match the triggering conditions with the current drilling conditions. As a preferred approach, the process of adaptively updating the structural surface anomaly triggering threshold is as follows: One or more of the following are used as mechanical response characteristic statistics: mean, standard deviation, median, quantile, moving mean, or moving median; within each flow-while-drilling window, the structural surface anomaly trigger threshold is dynamically updated based on the mechanical response characteristic statistics of the current stable drilling section of the borehole, as shown in the following formula; ; In the formula, For the first Structural anomaly triggering threshold of normalized mechanical response characteristics , For the current stable drilling section The median or mean of the normalized mechanical response characteristics. For the corresponding standard deviation, interquartile range, or absolute median, , which is the threshold amplification factor, a dimensionless configurable parameter determined based on historical borehole structure surface identification statistics or engineering experience.

[0031] In this technical solution, by identifying and updating stable drilling sections within the currently drilled borehole, the baseline response of drilling parameters is dynamically determined using the mean, median, or sliding statistics of the parameters within the stable drilling section. This provides a continuously adaptive reference benchmark for the normalization of various mechanical response characteristics as the drilling progresses. Furthermore, multi-dimensional mechanical response characteristics such as energy per unit footage, mechanical specific energy, torque-to-thrust ratio, drilling rate residual, and energy mutation coefficient are converted into dimensionless relative values ​​based on the stable drilling section. This effectively eliminates the influence of different drilling conditions, different borehole depths, and gradual changes in surrounding rock mechanical properties on the absolute values ​​of mechanical response characteristics. This ensures that the normalized mechanical response characteristics have a unified comparison benchmark across the entire borehole depth, providing reliable quantitative input for the adaptive updating of subsequent structural surface anomaly trigger thresholds and the accurate determination of candidate trigger windows.

[0032] Step 5: Anomaly Triggering and Boundary Fine Positioning Window Activation; When the normalized mechanical response characteristics meet the anomaly triggering conditions, the flowing drilling window is marked as a candidate triggering window, and the boundary fine positioning window centered on the candidate triggering window is activated; As a preferred approach, the process of exception triggering and fine-grained boundary positioning window activation is as follows: S51: When the normalized mechanical response characteristics within the flow-while-drilling window meet the abnormal triggering conditions, the flow-while-drilling window is marked as a candidate triggering window; the candidate triggering window is used to complete the coarse detection of structural anomalies. The abnormality triggering conditions include: [the following conditions are met] for a continuous period of not less than [number missing]. Within a given depth sampling point, at least one of the following: energy mutation coefficient, drilling rate residual, drilling stiffness change rate, peak mechanical energy, or torque-to-thrust ratio change... The characteristics exceed the corresponding structural surface anomaly trigger threshold; among them, and As a preset positive integer, The typical value is 3 to 5. The typical value is 2 to 3; S52: Activate the boundary fine positioning window centered on the candidate trigger window; the boundary fine positioning window is used to output the probability distribution of the starting boundary and the probability distribution of the ending boundary of the structural surface; the depth length of the boundary fine positioning window is greater than the depth length of the candidate trigger window, and its extension range extends to the front and rear sides of the candidate trigger window by a preset depth distance to ensure complete coverage of the boundary transition area of ​​the structural surface; In this technical solution, the mean, standard deviation, median, or slip statistic of the current stable drilling section of the borehole is used as the mechanical response characteristic statistic. Within each flowing-while-drilling window, the structural anomaly trigger threshold is dynamically updated. The trigger threshold is adaptively adjusted as the drilling progresses by using a formula: the median or mean plus a threshold amplification factor multiplied by the standard deviation, interquartile range, or absolute median. This adaptive mechanism ensures that the threshold setting precisely matches the actual drilling conditions and surrounding rock mechanical properties of the current borehole, effectively avoiding false alarms and missed alarms caused by fixed thresholds under different geological conditions and drilling parameters. This provides a reliable triggering benchmark for the accurate determination of subsequent candidate trigger windows.

[0033] Step 6: Model Inference and Multi-Probability Output; Input the drilling parameter sequence and normalized mechanical response features within the boundary fine-tuning window into the drilling structure surface identification model, and output the structure surface type probability, the starting boundary probability distribution, the ending boundary probability distribution, and the structure surface existence probability. As a preferred embodiment, the structure surface type probability is used to characterize the probability that the current boundary fine-positioning window belongs to different structure surface types, and the sum of the probabilities of each type is 1; the starting point boundary probability distribution is used to characterize the probability that the starting depth of the structure surface is at each depth sampling point within the boundary fine-positioning window, and the depth position corresponding to the probability peak is the most likely starting depth; the ending point boundary probability distribution is used to characterize the probability that the ending depth of the structure surface is at each depth sampling point within the boundary fine-positioning window, and the depth position corresponding to the probability peak is the most likely ending depth; the structure surface existence probability is used to characterize the probability that a structure surface exists within the current boundary fine-positioning window; all outputs are dimensionless probability values, with a value range of [0,1]; In this technical solution, the drilling parameter sequence and normalized mechanical response features within the boundary fine-tuning window are input into a drilling-only input model calibrated using borehole inspection data. Simultaneously, four types of probabilistic information are output: structure type probability, starting boundary probability distribution, ending boundary probability distribution, and structure existence probability. The structure type probability represents the structure category of the current window, with the sum of all probabilities equal to 1. The starting and ending boundary probability distributions characterize the uncertainty of the starting and ending boundary positions of the structure using probability values ​​at each depth sampling point, with the peak probability corresponding to the most probable boundary depth. This multi-probability joint output mechanism elevates structure identification from traditional discrete point localization to continuous probability representation, providing complete and directly computable probabilistic input for subsequent multi-window probability fusion, precise boundary localization, and identification confidence quantification.

[0034] Step 7: Multi-window probabilistic fusion and structural surface parameter determination; perform probabilistic fusion on the output results of multiple consecutive boundary fine-localization windows to determine the structural surface type, starting depth, ending depth and development length of the same structural surface candidate interval; As a preferred option, the process of multi-window probability fusion and structural surface parameter determination is as follows: S71: As Figure 3 As shown, when the structural surface types output by adjacent boundary fine positioning windows have the same maximum probability category, and the probability distributions of the starting boundary or the ending boundary overlap in the depth direction, the adjacent boundary fine positioning windows are merged into the same structural surface candidate interval. The overlap determination condition is: the probability distribution of the end boundary of the previous window and the probability distribution of the start boundary of the next window intersect in the depth coordinate, and the probability values ​​of both in the overlapping area are greater than the preset overlap probability threshold. S72: Weighted fusion of the probability distributions of the starting boundaries of multiple fine-positioning windows within the same candidate interval of the structural surface, and determination of the starting depth of the structural surface by the peak probability or expected probability position after fusion; weighted fusion of the probability distributions of the ending boundaries, and determination of the ending depth of the structural surface by the peak probability or expected probability position after fusion; the development length of the structural surface is obtained by subtracting the starting depth of the structural surface from the ending depth of the structural surface. The weights of the weighted fusion are determined based on one or more of the following: the probability of the structural surface type of the boundary fine-positioning window, the intensity of the physical response anomaly, the distance between the center depth of the window and the center depth of the candidate interval, and the completeness rate of the drilling parameters within the window. The larger the weight value, the more reliable the identification result of the window. The intensity of the physical response anomaly is the sum or peak value of the normalized mechanical response features within the boundary fine-positioning window that exceed the corresponding trigger threshold. In this technical solution, by judging the consistency of structural surface types between adjacent windows and the overlap of boundary probability distributions, continuous boundary fine-positioning windows are merged into the same structural surface candidate interval, effectively overcoming the problem of discontinuous boundary positioning caused by local noise interference in single-window recognition results. The boundary probability distributions of multiple windows within the same candidate interval are weighted and fused. The weights integrate multi-dimensional reliability indicators such as structural surface type probability, physical response anomaly intensity, window position distance, and drilling parameter integrity rate, ensuring that the recognition results of different windows participate in the fusion calculation according to their reliability. Finally, the starting and ending depths of the structural surface are determined by the peak probability or expected position after fusion, significantly improving the accuracy and robustness of structural surface boundary positioning.

[0035] Step 8: Calculate the recognition confidence and mark the areas to be reviewed; calculate the recognition confidence based on multi-dimensional information and output the structural surface recognition results; when the recognition confidence is lower than the preset review threshold, mark the corresponding depth interval as the area to be drilled for inspection and review. As a preferred approach, the process of identifying confidence level calculations and verification markers is as follows: S81: Calculate the identification confidence level based on the probability of structural surface type, the peak probability of the starting boundary, the peak probability of the ending boundary, the consistency of the continuous window, and the intensity of physical response anomalies. ,in The base confidence level is determined by the probability of the structural surface type, the peak probability of the starting boundary, and the peak probability of the ending boundary, and its value ranges from [0,1]. The window stability correction factor is determined by the consistency of continuous windows, and its value ranges from [0,1]. It is determined based on the consistency of the structural surface type output by adjacent windows within the same structural surface candidate interval, the overlap of the probability distribution of the starting boundary, and the overlap of the probability distribution of the ending boundary. This is a physical support correction factor determined by the intensity of the physical response anomaly, with a value range of [0,1]. It is determined based on the normalized value of the physical response anomaly intensity; the greater the anomaly intensity, the better. The closer to 1; S82: When the identification confidence level is lower than the preset verification threshold, or the peak width of the probability distribution of the starting point boundary and the probability distribution of the ending point boundary is greater than the preset uncertainty threshold, the corresponding depth interval is marked as the borehole inspection verification interval; the peak width is the depth interval width in the probability distribution where the probability value is greater than the preset probability ratio threshold (e.g., 0.5 times the peak value); S83: When the deviation between the subsequent borehole inspection verification result and the structure surface identification result in terms of structure surface type, structure surface start depth, or structure surface end depth exceeds a preset deviation threshold, the drilling parameter sequence and verification result within the corresponding depth interval are added to the incremental sample library for subsequent incremental training or parameter updates of the model.

[0036] This technical solution constructs an identification confidence calculation model by integrating basic confidence, continuous window consistency correction factors, and physical response anomaly intensity correction factors, achieving a multi-dimensional quantitative assessment of the reliability of structural surface identification results. When the identification confidence is lower than a preset verification threshold or the peak width of the boundary probability distribution is too large, the corresponding depth interval is automatically marked as a borehole inspection verification interval, providing clear guidance for verification priorities in field operations. Simultaneously, when the deviation between subsequent borehole inspection verification results and identification results exceeds a preset threshold, the verification data is automatically incorporated into the incremental sample library for incremental model training or parameter updates, forming a continuous improvement closed loop of identification, evaluation, verification, and optimization.

[0037] Example 2:

[0038] like Figure 2 As shown, this embodiment provides a real-time identification system for surrounding rock structural surfaces based on peep calibration during drilling, used to implement a method for real-time identification of surrounding rock structural surfaces based on peep calibration during drilling, including: The calibration run package calling module is used to call the drilling structural surface identification run package, which is pre-calibrated using borehole inspection data. The drilling structural surface identification run package is formed during the model training or calibration phase by using borehole inspection data to determine the structural surface type, starting depth, and ending depth, and by combining the drilling parameter sequence and mechanical response characteristics within the corresponding depth range to train and obtain a drilling-only input model calibrated using borehole inspection data. It includes at least the following parameters: drilling structural surface identification model parameters, standardized parameters for drilling parameter features, stable drilling section determination parameters, anomaly trigger initial threshold, boundary fine-positioning window parameters, multi-window fusion parameters, and confidence calculation parameters. The borehole inspection data is only used during the training, calibration, or verification phases; borehole inspection data is not input in real-time during the field identification phase. During the field identification phase, this module loads the run package into the field operation host, edge computing terminal, or drilling rig data processing unit, providing model parameters and configuration parameters for subsequent modules. The real-time acquisition module for drilling parameters is used to acquire drilling parameters of the borehole to be identified in real time during the on-site drilling process. The drilling parameters include multiple parameters such as drilling depth, drilling time, drilling speed, rotational speed, thrust, torque, drill pipe displacement, hydraulic pressure, drilling vibration, or drilling rig power. During the on-site identification stage, only the drilling parameters acquired in real time during the drilling process are used to complete the structural surface identification, without using borehole peep images, borehole videos, or borehole unfolding diagrams as real-time input. The streaming window management module is used to divide the real-time acquired drilling parameters into continuous streaming drilling windows according to a preset depth step. Each streaming drilling window corresponds to a continuous drilling depth range. In the current drilled section, the module identifies stable drilling sections where the mechanical response fluctuation amplitude is lower than a preset stability threshold and is not marked as a candidate section of the structural surface. When the normalized mechanical response characteristics in the streaming drilling window meet the abnormal triggering conditions, the module marks the streaming drilling window as a candidate triggering window and starts a boundary fine positioning window that extends a preset depth distance to the front and back sides of the candidate triggering window. The stable section benchmark update module is used to update the drilling parameter benchmark response online based on the stable drilling section in the current drilled section. The drilling parameter benchmark response is determined according to the mean, median, standard deviation, quantile or slip statistic of the drilling parameters in the stable drilling section. As the drilling progresses and new stable drilling sections are identified, the drilling parameter benchmark response is continuously updated online to adapt to the gradual change of the surrounding rock mechanical properties in different depth sections, and to provide a dynamic reference benchmark for the calculation of normalized mechanical response characteristics. The mechanical response characteristic calculation module is used to calculate the normalized mechanical response characteristics within the flow-while-drilling window. The normalized mechanical response characteristics include multiple parameters such as energy per unit footage, mechanical specific energy, torque-to-thrust ratio, drilling speed residual, and energy mutation coefficient. All parameters are based on the reference response of the drilling parameters in the stable drilling section. Each mechanical response characteristic is converted to a dimensionless relative value based on the stable drilling section to eliminate the influence of different drilling conditions and borehole depths on the absolute values ​​of the mechanical response characteristics. The adaptive threshold update module is used to dynamically update the structural surface anomaly trigger threshold within each flow-while-drilling window, using the mean, standard deviation, median, quantile, moving mean, or moving median as statistical features of the mechanical response. Normalized mechanical response characteristics, and their structural surface anomaly triggering threshold ,in This represents the median or mean of this characteristic within the current stable drilling section of the borehole. For the corresponding standard deviation, interquartile range, or absolute median, This is the threshold amplification factor; this module enables the trigger threshold to be adaptively adjusted as the drilling progresses, precisely matching the actual drilling conditions and surrounding rock mechanical properties of the current borehole. The boundary probability inference module is used to input the drilling parameter sequence and normalized mechanical response features within the boundary fine-positioning window into the loaded drilling structure identification model, and outputs the structure type probability, the starting boundary probability distribution, the ending boundary probability distribution, and the structure existence probability. The structure type probability is used to characterize the probability that the current window belongs to different structure types, and the sum of the probabilities of each type is 1. The starting boundary probability distribution and the ending boundary probability distribution are used to characterize the probability of the starting depth and ending depth of the structure at each depth sampling point within the window, respectively. The structure existence probability is used to characterize the probability that a structure exists within the current window. All outputs are dimensionless probability values, with a value range of [0,1]. The multi-window fusion module is used to probabilistically fuse the output results of multiple consecutive boundary fine positioning windows. When the maximum probability category of the structural surface type output by adjacent windows is the same, and the probability distribution of the starting boundary or the probability distribution of the ending boundary overlaps in the depth direction, the adjacent windows are merged into the same structural surface candidate interval. The probability distributions of the starting boundary and the probability distribution of the ending boundary of multiple windows in the same candidate interval are weighted and fused respectively. The weights are determined comprehensively based on the probability of the structural surface type of the window, the intensity of the physical response anomaly, the distance between the center depth of the window and the center depth of the candidate interval, and the drilling parameter integrity rate within the window. The starting depth and ending depth of the structural surface are determined by the peak probability or the expected probability position after fusion, and the development length of the structural surface is obtained by subtracting the starting depth from the ending depth. The confidence and verification labeling module is used to calculate and identify confidence based on the probability of structural surface type, the peak probability of the starting boundary, the peak probability of the ending boundary, the consistency of the continuous window, and the intensity of physical response anomalies. ,in Based on the base confidence level, take the weighted geometric mean or weighted arithmetic mean of the peak probabilities; This is a continuous window consistency correction factor; This is a physical response anomaly intensity correction factor; when the identification confidence is lower than the preset verification threshold, or the peak width of the boundary probability distribution is greater than the preset uncertainty threshold, the corresponding depth interval is marked as the borehole inspection verification interval; The incremental sample management module is used to add the drilling parameter sequence and verification results within the corresponding depth range to the incremental sample library when the deviation between the structure type, starting depth or ending depth between the subsequent borehole inspection verification results and the structure identification results exceeds the preset deviation threshold. This is used for incremental training or parameter updates of the subsequent model, forming a continuous improvement closed loop of identification, evaluation, verification and optimization.

[0039] Based on a modular collaborative architecture, this system realizes the complete transformation from high-precision calibration of surrounding rock structural surfaces through borehole inspection to real-time and efficient identification of parameters while drilling. It can effectively solve the problems in existing technologies, such as the difficulty of real-time application of borehole inspection, the significant impact of working condition fluctuations on independent identification while drilling, insufficient probabilistic output of structural surface boundaries, and the lack of verification and marking mechanisms for low-confidence identification results. It provides a reliable, efficient, and evolvable system-level solution for rapid on-site detection and intelligent identification of surrounding rock structural surfaces in underground engineering.

[0040] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for real-time identification of surrounding rock structural surfaces during drilling based on observation calibration, characterized in that, Includes the following steps: Step 1: Obtain the drilling structural surface identification operation package pre-calibrated using borehole inspection data; calibrate the drilling structural surface identification model using borehole inspection data during the model training or calibration phase to form an operation package that relies solely on drilling parameters during the field identification phase. Step 2: Real-time acquisition of drilling parameters and construction of a flow-based drilling window; During the drilling process on site, real-time acquisition of drilling parameters, construction of a flow-based drilling window according to the preset depth step size, and determination of stable drilling sections from the drilled sections; Step 3: Update the baseline response of the drilling parameters online and calculate the normalized mechanical response characteristics; Based on the stable drilling section in the current drilled section, the baseline response of the drilling parameters is updated online, and the normalized mechanical response characteristics within the flow-while-drilling window are calculated. Step 4: Adaptively update the structural surface anomaly trigger threshold; dynamically update the structural surface anomaly trigger threshold for each normalized mechanical response feature based on the mechanical response characteristic statistics of the current stable drilling section of the borehole. Step 5: Anomaly Triggering and Boundary Fine Positioning Window Activation; When the normalized mechanical response characteristics meet the anomaly triggering conditions, the flowing drilling window is marked as a candidate triggering window, and the boundary fine positioning window centered on the candidate triggering window is activated; Step 6: Model Inference and Multi-Probability Output; Input the drilling parameter sequence and normalized mechanical response features within the boundary fine-tuning window into the drilling structure surface identification model, and output the structure surface type probability, the starting boundary probability distribution, the ending boundary probability distribution, and the structure surface existence probability. Step 7: Multi-window probabilistic fusion and structural surface parameter determination; perform probabilistic fusion on the output results of multiple consecutive boundary fine-localization windows to determine the structural surface type, starting depth, ending depth and development length of the same structural surface candidate interval; Step 8: Identify confidence level calculations and markers to be reviewed; The identification confidence level is calculated based on multi-dimensional information, and the structural surface identification result is output. When the identification confidence level is lower than the preset verification threshold, the corresponding depth interval is marked as the borehole inspection verification interval.

2. The method for real-time identification of surrounding rock structural surfaces based on observation calibration according to claim 1, characterized in that, In step 1, the process of acquiring the pre-calibrated drilling structure surface identification runtime package formed by borehole inspection data is as follows: During the model training or calibration phase, borehole inspection data is used to determine the structure type, the starting depth of the structure, and the ending depth of the structure. Combined with the drilling parameter sequence and mechanical response characteristics within the corresponding depth range, a drilling-only input model calibrated by borehole inspection data is trained and obtained as the drilling structure identification model. The drilling structure identification model uses only the drilling parameter sequence and normalized mechanical response characteristics as input during the field identification phase. After training or calibration is completed, the model parameters, standardized parameters of drilling parameters, stable drilling section judgment parameters, abnormal trigger initial threshold, boundary fine positioning window parameters, multi-window fusion parameters and confidence calculation parameters are packaged into a drilling structure surface recognition runtime package. The drilling structure identification runtime package includes at least the drilling structure identification model parameters, drilling parameter feature standardization parameters, stable drilling section determination parameters, abnormal trigger initial threshold, boundary fine positioning window parameters, multi-window fusion parameters, and confidence calculation parameters; wherein, the borehole inspection data is only used in the training, calibration, or verification phases to determine the structure type, structure start depth, and structure end depth.

3. The method for real-time identification of surrounding rock structural surfaces based on observation calibration according to claim 1, characterized in that, In step 2, the process of acquiring drilling parameters in real time and constructing a streaming drilling window is as follows: S21: During the on-site drilling process, the drilling parameters of the borehole to be identified are collected in real time; the drilling parameters include multiple parameters such as drilling depth, drilling time, drilling speed, rotation speed, thrust, torque, drill pipe displacement, hydraulic pressure, drilling vibration, or drilling rig power; the on-site identification stage only uses the drilling parameters collected in real time during the drilling process to complete the structural surface identification; S22: Load the drilling structure surface identification and operation package into the field operation host, edge computing terminal or drilling rig data processing unit; divide the drilling parameters into continuous flow drilling windows according to the preset depth step size, and each flow drilling window corresponds to a continuous drilling depth range; S23: In the currently drilled section, if the mechanical response fluctuation amplitude in a certain depth section is lower than the preset stability threshold and is not marked as a candidate section for structural surface, then it is determined as a stable drilling section; the mechanical response fluctuation amplitude is determined based on the coefficient of variation or standard deviation of at least one of the parameters of drilling speed, rotation speed, thrust and torque within the depth window.

4. The method for real-time identification of surrounding rock structural surfaces based on observation calibration according to claim 1, characterized in that, In step 3, the process of updating the baseline response of the drilling parameters online and calculating the normalized mechanical response characteristics is as follows: S31: Based on the stable drilling section in the current drilled section, the baseline response of the drilling parameters is updated online; the baseline response of the drilling parameters is determined according to the mean, median, standard deviation, quantile or slip statistic of the drilling parameters in the stable drilling section; as the drilling progresses and new stable drilling sections are identified, the baseline response of the drilling parameters is continuously updated online to adapt to the gradual change of the surrounding rock mechanical properties in different depth sections; S32: Calculate the normalized mechanical response characteristics within the flow-while-drilling window; the normalized mechanical response characteristics include multiple of the following: energy per unit footage, mechanical specific energy, torque-to-thrust ratio, drilling rate residual, and energy mutation coefficient; wherein, energy per unit footage is the ratio of drilling energy to the corresponding footage; mechanical specific energy is the energy required to break a unit volume of rock; torque-to-thrust ratio is the ratio of torque to thrust; drilling rate residual is determined based on the difference between the measured drilling rate and the reference drilling rate predicted from the current thrust, rotational speed, and torque; and energy mutation coefficient is determined based on the normalized deviation of the energy per unit footage within the current flow-while-drilling window from the reference value of energy per unit footage in the stable drilling section. Among them, each of the above-mentioned normalized mechanical response characteristics is based on the reference response of the drilling parameters in the stable drilling section, and each mechanical response characteristic is converted to a dimensionless relative value based on the stable drilling section.

5. The method for real-time identification of surrounding rock structural surfaces based on observation calibration according to claim 1, characterized in that, In step 4, the process of adaptively updating the structural surface anomaly triggering threshold is as follows: One or more of the following are used as mechanical response characteristic statistics: mean, standard deviation, median, quantile, moving mean, or moving median; within each flow-while-drilling window, the structural surface anomaly trigger threshold is dynamically updated based on the mechanical response characteristic statistics of the current stable drilling section of the borehole, as shown in the following formula; ； In the formula, For the first Structural anomaly triggering threshold of normalized mechanical response characteristics , For the current stable drilling section The median or mean of the normalized mechanical response characteristics. For the corresponding standard deviation, interquartile range, or absolute median, This is the threshold amplification factor.

6. The method for real-time identification of surrounding rock structural surfaces based on observation calibration according to claim 1, characterized in that, In step 5, the process of exception triggering and starting the fine-grained boundary positioning window is as follows: S51: When the normalized mechanical response characteristics within the flow-while-drilling window meet the abnormal triggering conditions, the flow-while-drilling window is marked as a candidate triggering window. The abnormality triggering conditions include: [the following conditions are met] for a continuous period of not less than [number missing]. Within a given depth sampling point, at least one of the following: energy mutation coefficient, drilling rate residual, drilling stiffness change rate, peak mechanical energy, or torque-to-thrust ratio change... The characteristic exceeds the corresponding structural surface anomaly trigger threshold; S52: Activate the boundary fine positioning window centered on the candidate trigger window; the boundary fine positioning window is used to output the probability distribution of the starting boundary and the probability distribution of the ending boundary of the structural surface; the depth length of the boundary fine positioning window is greater than the depth length of the candidate trigger window, and its extension range extends to the front and rear sides of the candidate trigger window by a preset depth distance to ensure complete coverage of the boundary transition area of ​​the structural surface.

7. The method for real-time identification of surrounding rock structural surfaces based on observation calibration according to claim 1, characterized in that, In step 6, the structure surface type probability is used to characterize the probability that the current boundary fine positioning window belongs to different structure surface types, and the sum of the probabilities of each type is 1; The starting point boundary probability distribution is used to characterize the probability of the starting depth of the structural surface at each depth sampling point within the boundary fine positioning window, and the depth position corresponding to the probability peak is the most likely starting depth; the ending point boundary probability distribution is used to characterize the probability of the ending depth of the structural surface at each depth sampling point within the boundary fine positioning window, and the depth position corresponding to the probability peak is the most likely ending depth; the structural surface existence probability is used to characterize the probability of the existence of a structural surface within the current boundary fine positioning window; each output is a dimensionless probability value with a value range of [0,1].

8. The method for real-time identification of surrounding rock structural surfaces based on observation calibration according to claim 1, characterized in that, In step 7, the process of multi-window probability fusion and structural surface parameter determination is as follows: S71: When the structural surface type output by adjacent boundary fine positioning windows has the same maximum probability category, and the probability distribution of the starting boundary or the probability distribution of the ending boundary overlaps in the depth direction, the adjacent boundary fine positioning windows are merged into the same structural surface candidate interval. The overlap determination condition is: the probability distribution of the end boundary of the previous window and the probability distribution of the start boundary of the next window intersect in the depth coordinate, and the probability values ​​of both in the overlapping area are greater than the preset overlap probability threshold. S72: Weighted fusion of the probability distributions of the starting boundaries of multiple fine-positioning windows within the same candidate interval of the structural surface, and determination of the starting depth of the structural surface by the peak probability or expected probability position after fusion; weighted fusion of the probability distributions of the ending boundaries, and determination of the ending depth of the structural surface by the peak probability or expected probability position after fusion; the development length of the structural surface is obtained by subtracting the starting depth of the structural surface from the ending depth of the structural surface. The weights of the weighted fusion are determined based on one or more of the following: the probability of the structural surface type of the boundary fine-positioning window, the intensity of the physical response anomaly, the distance between the center depth of the window and the center depth of the candidate interval, and the completeness rate of the drilling parameters within the window. The larger the weight value, the more reliable the identification result of the window. The intensity of the physical response anomaly is the sum or peak value of the normalized mechanical response features within the boundary fine-positioning window that exceed the corresponding trigger threshold.

9. The method for real-time identification of surrounding rock structural surfaces based on observation calibration according to claim 1, characterized in that, In step 8, the process of identifying the confidence level calculation and the markers to be reviewed is as follows: S81: Calculate the identification confidence level based on the probability of structural surface type, the peak probability of the starting boundary, the peak probability of the ending boundary, the consistency of the continuous window, and the intensity of physical response anomalies. ,in The base confidence level is determined by the probability of the structural surface type, the peak probability of the starting boundary, and the peak probability of the ending boundary. This is the window stability correction factor determined by the consistency of continuous windows; This is a physical support correction factor determined by the intensity of the physical response anomaly. S82: When the identification confidence level is lower than the preset verification threshold, or the peak width of the probability distribution of the starting point boundary and the probability distribution of the ending point boundary is greater than the preset uncertainty threshold, the corresponding depth interval is marked as the borehole inspection verification interval; the peak width is the depth interval width in the probability distribution where the probability value is greater than the preset probability ratio threshold; S83: When the deviation between the subsequent borehole inspection verification result and the structure surface identification result in terms of structure surface type, structure surface start depth, or structure surface end depth exceeds a preset deviation threshold, the drilling parameter sequence and verification result within the corresponding depth interval are added to the incremental sample library for subsequent incremental training or parameter updates of the model.

10. A real-time identification system for surrounding rock structural surfaces based on peep calibration during drilling, used to implement the real-time identification method for surrounding rock structural surfaces based on peep calibration during drilling as described in any one of claims 1 to 9, characterized in that, include: The calibration operation package calling module is used to call the drilling structure surface identification operation package that has been pre-calibrated using borehole inspection data; The real-time acquisition module for drilling parameters is used to acquire the drilling parameters of the borehole to be identified in real time during the drilling process on site. The streaming window management module is used to divide the real-time acquired drilling parameters into continuous streaming drilling windows according to a preset depth step size. Each streaming drilling window corresponds to a continuous drilling depth range. The stable section benchmark update module is used to update the benchmark response of drilling parameters online based on the stable drilling section in the current drilled section of the borehole. The mechanical response characteristic calculation module is used to calculate the normalized mechanical response characteristics within the flow-while-drilling window; The adaptive threshold update module is used to dynamically update the structural surface anomaly trigger threshold within each flow-while-drilling window based on the statistics of the current stable drilling section of the borehole, using the mean, standard deviation, median, quantile, moving mean, or moving median as the mechanical response characteristic statistics. The boundary probability inference module is used to input the drilling parameter sequence and normalized mechanical response features within the boundary fine positioning window into the loaded drilling structure surface identification model, and output the structure surface type probability, the starting boundary probability distribution, the ending boundary probability distribution, and the structure surface existence probability. The multi-window fusion module is used to probabilistically fuse the output results of multiple consecutive boundary fine-positioning windows; The confidence and verification labeling module is used to calculate and identify confidence based on the probability of structural surface type, the peak probability of the starting boundary, the peak probability of the ending boundary, the consistency of the continuous window, and the intensity of physical response anomalies. When the identification confidence level is lower than the preset verification threshold, or the peak width of the boundary probability distribution is greater than the preset uncertainty threshold, the corresponding depth interval is marked as the borehole inspection verification interval. The incremental sample management module is used to add the drilling parameter sequence and verification results within the corresponding depth range to the incremental sample library when the deviation between the structure type, starting depth or ending depth between the subsequent borehole inspection verification results and the structure identification results exceeds a preset deviation threshold.

Images