Rock mass mechanics parameter measuring model construction method and while-drilling measuring method

Abstract

This invention relates to the field of geological technology, and provides a method for constructing a rock mass mechanics parameter measurement model and a method for measuring rock mass mechanics parameters while drilling. The rock mass mechanics parameter measurement model constructed by the method is used to predict rock mass mechanics parameters based on a real-time collected basic sample set and corresponding second extended sample subsets. It can realize real-time measurement of rock mass mechanics parameters during drilling. Moreover, during the training process, features representing rock mass mechanics knowledge-driven features are introduced as input parameters for the rock mass mechanics parameter measurement model. Furthermore, through a depth sliding window mechanism, statistical feature parameters related to the local microscopic characteristics, transitional characteristics, and overall macroscopic characteristics of the rock mass are added as input parameters to train the rock mass mechanics parameter measurement model, so that the trained rock mass mechanics parameter measurement model can more accurately predict rock mass mechanics parameters.

Description

Technical Field

[0001] This invention relates to the field of rock mass engineering technology, and in particular to a method for constructing a model for measuring rock mass mechanical parameters and a method for measuring them while drilling. Background Technology

[0002] In rock engineering practice, the accurate acquisition of rock mass mechanical parameters (such as uniaxial compressive strength (UCS), elastic modulus (E), and fault fracture zone distribution) is a core prerequisite for ensuring construction safety and efficiency. However, existing technologies mainly employ the following two methods to obtain these rock mass mechanical parameters: Post-drilling acquisition methods: Traditional methods for obtaining rock mass mechanical parameters mainly rely on core sampling after drilling or borehole television observation with lag. The former requires sending core samples to a laboratory for mechanical testing (such as UCS testing) after drilling is completed, a process that takes ≥24 hours, with sampling intervals typically between 5 and 10 meters, failing to reflect small-scale changes in rock mass parameters. The latter, while allowing for rapid observation of rock mass structure (such as fault fracture zones) after drilling, still requires lowering the observation equipment after drilling stops, resulting in a lag time of ≥5 hours. This lag prevents real-time early warning of engineering risks.

[0003] The method of obtaining parameters while drilling focuses only on acquiring raw mechanical parameters such as drilling displacement, rotational speed, and thrust, without introducing rock mass mechanical characteristic indicators, which are strongly correlated with the hardness and drillability of the rock mass. Due to the lack of such "mechanical knowledge-driven" characteristic indicators, the accuracy of rock mass mechanical parameter prediction in existing technologies is low and cannot meet the requirements of engineering for parameter accuracy. Summary of the Invention

[0004] This invention provides a method for constructing a rock mass mechanical parameter determination model and a method for measuring parameters while drilling, in order to solve the above-mentioned technical problems existing in the prior art.

[0005] This invention provides a method for constructing a rock mass mechanical parameter determination model, comprising the following steps: Multiple sets of mechanical parameter samples are obtained during the drilling process, and each set of mechanical parameter samples includes multiple mechanical parameter samples; Based on each set of mechanical parameter samples and rock mechanics theory, a set of reconstructed parameter samples characterizing the mechanical properties of rock mass is determined to obtain a basic sample set consisting of multiple sets of mechanical parameter samples and multiple sets of reconstructed parameter samples. Based on multiple depth sliding windows of different scales, m statistical feature parameters of each of the n basic samples in each depth sliding window of each scale are calculated to obtain multiple extended sample sets. Each extended sample set includes multiple extended sample groups, and each extended sample group includes n×m statistical feature parameters, where n and m are both greater than or equal to 1. For each of the extended sample sets, the statistical feature parameters in the extended sample set are optimized based on the correlation between each statistical feature parameter to obtain the optimized first extended sample subset; Obtain the rock mass mechanics parameter labels corresponding to the basic sample set and each first extended sample subset; For each of the first extended sample subsets, based on the importance of each statistical feature parameter to the rock mechanics parameter label, the statistical feature parameters in each of the first extended sample subsets are optimized to obtain each of the optimized second extended sample subsets; Based on the basic sample set, each second extended sample subset, and their corresponding rock mass mechanics parameter labels, a rock mass mechanics parameter measurement model is trained. The rock mass mechanics parameter measurement model is used to predict rock mass mechanics parameters based on the real-time collected basic sample set and the corresponding second extended sample subsets.

[0006] According to the method for constructing a rock mass mechanical parameter measurement model provided by the present invention, multiple sets of mechanical parameter sample groups are obtained during the drilling process, including: Obtain multiple sets of original mechanical parameters during the drilling process; Based on the drilling depth, drilling pressure, and drilling speed in each original mechanical parameter set, as well as the constraints of the stable drilling stage, determine whether each original mechanical parameter set corresponds to the stable drilling stage. The original set of mechanical parameters corresponding to the stable drilling stage is determined as the mechanical parameter sample set.

[0007] According to the rock mass mechanics parameter determination model construction method provided by the present invention, before determining the reconstructed parameter sample set characterizing the rock mass mechanics properties based on each set of mechanical parameter sample sets and rock mass mechanics theory, the method further includes: The drilling speed in the mechanical parameter sample group is denoised using a Kalman filter algorithm. The drilling pressure and drill rod torque in the mechanical parameter sample group were denoised using a wavelet threshold denoising algorithm. The drilling current in the mechanical parameter sample group is denoised using an adaptive noise complete set empirical mode decomposition algorithm.

[0008] According to the rock mass mechanics parameter determination model construction method provided by the present invention, based on each set of mechanical parameter sample groups and rock mass mechanics theory, a reconstructed parameter sample group characterizing the rock mass mechanics properties is determined, including: Based on the drilling pressure, drilling speed, drill rod torque and drilling speed in the mechanical parameter sample group, and combined with the rock mechanics theory, the drilling specific energy characterizing the hardness of the rock mass is determined. Based on the average values ​​of drilling pressure, drilling speed, and drilling rotation speed in the mechanical parameter sample group at the preset drilling displacement, and combined with the rock mechanics theory, the rock mass strength correlation coefficient is determined. Based on the drilling pressure, drilling speed, drilling rotation speed and the correlation coefficient of rock mass strength in the mechanical parameter sample group, and combined with the rock mechanics theory, the drilling power index characterizing the drillability of the rock mass is determined. A set of reconstructed parameter samples is formed using the drilling specific energy, rock mass strength correlation coefficient, and drilling power index.

[0009] According to the rock mass mechanics parameter determination model construction method provided by the present invention, for each of the extended sample sets, based on the correlation between each statistical characteristic parameter, the statistical characteristic parameters in the extended sample set are optimized to obtain an optimized first extended sample subset, including: For any of the extended sample sets, calculate the variance of each statistical feature parameter in each extended sample group, and delete statistical feature parameters whose variance is less than a preset variance threshold. Calculate the correlation coefficient between each pair of statistical feature parameters among the remaining statistical feature parameters. If the correlation coefficient is greater than the preset correlation coefficient threshold, retain one of the pair of statistical feature parameters to obtain the first extended sample subset.

[0010] According to the method for constructing a rock mass mechanics parameter determination model provided by the present invention, the rock mass mechanics parameter labels corresponding to the basic sample set and each first extended sample subset are obtained, including: Obtain the true values ​​of rock mass mechanical parameters as a function of depth obtained from core sampling after drilling is completed; The true values ​​of the rock mass mechanical parameters are interpolated according to a preset minimum depth step to obtain the rock mass mechanical parameter labels corresponding to each basic sample group in the basic sample set. The mechanical parameter sample groups in each basic sample group are obtained based on the preset minimum depth step. After interpolation, the corresponding rock mechanics parameter labels are associated with the base sample group in the base sample set and the extended sample group in each extended sample set according to depth.

[0011] According to the rock mass mechanics parameter determination model construction method provided by the present invention, for each first extended sample subset, based on the importance of each statistical feature parameter to the rock mass mechanics parameter label, the statistical feature parameters in each first extended sample subset are optimized to obtain optimized second extended sample subsets, including: For any first extended sample subset, based on each extended sample group and the corresponding rock mechanics parameter label in the first extended sample subset, a pre-trained model is trained and validated to obtain the feature importance score of each statistical feature parameter output by the pre-trained model. The features are sorted from largest to smallest according to their importance scores, and the feature importance scores are accumulated according to the sorting. When the accumulated score exceeds the score threshold, the accumulation stops, and the statistical feature parameters that participated in the accumulation in each of the expanded sample groups of any first expanded sample subset are retained to obtain the second expanded sample subset corresponding to any first expanded sample subset.

[0012] According to the present invention, a method for constructing a rock mass mechanics parameter determination model is provided, which trains the rock mass mechanics parameter determination model based on the basic sample set, each second extended sample subset, and their respective corresponding rock mass mechanics parameter labels, including: Based on the basic sample set, each second extended sample subset, and their corresponding rock mechanics parameter labels, multiple candidate models are trained, tested, and validated, and the candidate model with the best model stability index is selected as the rock mechanics parameter determination model.

[0013] This invention also provides a method for determining rock mass mechanical parameters while drilling, comprising the following steps: Obtain the real-time acquisition of the basic parameter set and the corresponding subsets of the second extended parameters; The basic parameter set and the corresponding second extended parameter subsets are respectively input into the rock mass mechanics parameter determination model constructed by the rock mass mechanics parameter determination model construction method described above, so as to obtain the basic parameter set and the candidate rock mass mechanics parameters corresponding to each of the second extended parameter subsets output by the rock mass mechanics parameter determination model. Based on the basic parameter set and the candidate rock mechanics parameters corresponding to each of the second extended parameter subsets, the final rock mechanics parameters are determined.

[0014] According to the present invention, a method for determining rock mass mechanics parameters while drilling, based on a basic parameter set and candidate rock mass mechanics parameters corresponding to each of the second extended parameter subsets, determines the final rock mass mechanics parameters, including: For rock mechanics parameters with continuous values, the deviation is determined based on the maximum and minimum candidate rock mechanics parameters. If the deviation is less than or equal to a preset deviation threshold, the final rock mechanics parameters are obtained by weighting each candidate rock mechanics parameter and its corresponding model test accuracy. Otherwise, after deleting the candidate rock mechanics parameter with the largest deviation, the final rock mechanics parameters are obtained by weighting the remaining candidate rock mechanics parameters and their corresponding model test accuracy. Here, the model test accuracy is the accuracy value of the basic sample set and each second extended sample subset during the test phase of model training. For rock mechanics parameters with Boolean values, the final rock mechanics parameter is determined from the candidate rock mechanics parameters based on the majority voting method. When the number of votes for each candidate rock mechanics parameter with different values ​​is equal, a weighted calculation is performed using the predicted probability of each candidate rock mechanics parameter with different values ​​and the corresponding F1 score as weights to obtain the comprehensive probability of different values. The value with the larger comprehensive probability is selected as the final value of the rock mechanics parameter. The F1 score is obtained by testing the basic sample set and each second extended sample subset during the testing phase of the model training process. Each sample set corresponds one-to-one with each parameter set, and each sample set corresponds to one F1 score.

[0015] The rock mass mechanics parameter measurement model construction method provided by this invention constructs a rock mass mechanics parameter measurement model, which is used to predict rock mass mechanics parameters based on the real-time collected basic sample set and the corresponding second extended sample subsets. It can realize the real-time measurement of rock mass mechanics parameters during drilling. Moreover, during the training process, features that characterize rock mass mechanics knowledge are introduced as input parameters of the rock mass mechanics parameter measurement model. Furthermore, through the depth sliding window mechanism, statistical feature parameters related to the local microscopic properties, transitional properties, and overall macroscopic properties of the rock mass are added as input parameters to train the rock mass mechanics parameter measurement model, so that the trained rock mass mechanics parameter measurement model can more accurately predict rock mass mechanics parameters. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the method for constructing a rock mass mechanical parameter determination model provided by the present invention.

[0018] Figure 2 This is a schematic flowchart of the method for determining rock mass mechanical parameters while drilling provided by the present invention.

[0019] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0021] The method for constructing a rock mass mechanical parameter determination model according to embodiments of the present invention, such as... Figure 1 As shown, it includes steps S110 to S170.

[0022] Step S110: Obtain multiple sets of mechanical parameter samples during the drilling process. Each set of mechanical parameter samples includes multiple mechanical parameter samples. These mechanical parameter samples include parameters such as drilling depth, drilling speed, drilling pressure, drill pipe torque, drilling rate, and drilling current. Each mechanical parameter sample is acquired through corresponding sensors installed on the drilling rig. For example, for commonly used geological drilling rigs, each mechanical parameter sample is acquired in the following way: Drilling depth: A laser displacement sensor (accuracy ±0.01mm, range 0~500mm) is installed next to the guide rail of the drilling rig to collect the displacement increment of the drill rod along the depth direction in real time, thereby obtaining the drilling depth.

[0023] Drilling speed: A Hall speed sensor (accuracy ±1r / min, range 0~300r / min) is used, coupled to the non-transmission end of the drill pipe drive shaft to avoid interference from transmission vibration on speed detection.

[0024] Drilling pressure: A piezoelectric force sensor (accuracy ±0.5%FS, range matched to the maximum thrust of the drilling rig) is integrated into the connecting flange between the drilling rig power head and the drill rod to directly obtain the axial thrust during the drilling process as the drilling pressure.

[0025] Drill pipe torque: The drill pipe torque is obtained by collecting the torsional load of the drill pipe through a drill pipe torque sensor (measurement accuracy ±0.2%FS, range 0~10kN·m).

[0026] Drilling current: The power head current sensor (measurement accuracy ±0.5%FS, range 0~50A) collects the power head operating current to obtain the drilling current.

[0027] Drilling speed: The real-time drilling speed is obtained by dividing the drilling depth by the time.

[0028] In this step, multiple sets of mechanical parameter samples can be obtained simply by acquiring the data collected by each sensor.

[0029] It should be noted that although the data acquisition frequencies of the aforementioned sensors are uniform, mechanical parameter sample groups at the same time can be obtained by aligning the timestamps. To match the depth distribution characteristics of the rock mass mechanical parameters, preferably, the acquired mechanical parameter sample groups are sampled and / or interpolated according to depth to obtain multiple mechanical parameter sample groups at preset depth intervals (e.g., 0.01m).

[0030] Step S120: Based on each set of mechanical parameter samples and rock mechanics theory, determine the reconstructed parameter sample set characterizing the rock mechanics properties to obtain a basic sample set consisting of multiple sets of mechanical parameter sample sets and multiple sets of reconstructed parameter sample sets. For example, at least one recognized characteristic index in the field of rock mechanics (e.g., drilling specific energy SE, drilling power index DPI, and rock strength correlation coefficient) is introduced as a reconstructed parameter sample to form a reconstructed parameter sample set. The basic sample set consists of multiple sets of mechanical parameter sample sets and multiple sets of reconstructed parameter sample sets. The basic sample set includes multiple basic sample sets, each consisting of one mechanical parameter sample set and a corresponding reconstructed parameter sample set. For example, the basic sample sets include: drilling depth, drilling speed, drilling pressure, drill pipe torque, drilling speed, drilling current, drilling specific energy SE, drilling power index DPI, and rock strength correlation coefficient. Each basic sample set includes 6 mechanical parameter samples and 3 reconstructed parameter samples, for a total of 9 basic samples.

[0031] By introducing features that characterize rock mechanics knowledge as input parameters to train the rock mechanics parameter measurement model, the trained model can more accurately predict rock mechanics parameters.

[0032] Step S130: Based on multiple depth sliding windows of different scales, calculate m statistical feature parameters for each of the n basic samples in each depth sliding window of each scale, to obtain multiple extended sample sets. Each extended sample set includes multiple extended sample groups, and each extended sample group includes n×m statistical feature parameters, where n and m are both greater than or equal to 1.

[0033] In existing technologies, the basic sample set obtained from the original collected data is directly used as the sample to train the model. This cannot take into account both the local microscopic characteristics of the rock mass (e.g., local fracturing) and the macroscopic characteristics of the strata (e.g., stratigraphic interfaces). As a result, the accuracy of the rock mass mechanical parameter measurement model trained in this way is relatively low. Therefore, in this embodiment, multiple depth sliding windows of different scales are designed so that different datasets focus on the "microscopic-transitional-macroscopic" characteristics of the rock mass. This provides more feature-rich sample datasets for subsequent training of the rock mass mechanical parameter measurement model, thereby enabling the trained rock mass mechanical parameter measurement model to predict rock mass mechanical parameters more accurately.

[0034] For example, based on the rock mass depth distribution characteristics, three scales of depth sliding windows are constructed, with the specific design as follows: Short-term depth sliding window: The window corresponds to a rock mass depth of 0.5m (i.e., the size of the depth sliding window is 0.5m, containing approximately 50 sampling points), and is used to extract the microscopic local characteristics of the rock mass (such as: local fractures, parameter abrupt changes caused by hard rock interlayers). Intermediate depth sliding window: The window corresponds to a rock mass depth of 2.5m (i.e., the size of the depth sliding window is 2.5m, containing approximately 250 sampling points), and is used to extract the transition characteristics of the rock mass (e.g., parameter gradation at the interface of different lithological strata). Long-term depth sliding window: The window corresponds to a rock mass depth of 5.0m (i.e., the size of the depth sliding window is 5m, containing approximately 500 sampling points), and is used to extract the overall macroscopic characteristics of the rock mass (e.g., the average rock mass mechanical parameters of the same stratum).

[0035] Taking a set of basic samples containing 9 basic samples as an example, i.e., n=9, for multiple sets of basic samples within each depth sliding window (e.g., a short-term depth sliding window has 50 sampling points, i.e., there are 50 sets of basic samples within one short-term depth sliding window), four types of statistical feature parameters (a total of 24 statistical feature parameters) are extracted as follows: Time-domain features (12 items): maximum value, minimum value, average value, peak-to-peak value, peak value, variance, standard deviation, kurtosis, skewness, root mean square value, shape factor, and peak factor.

[0036] Frequency domain features (5 items): Perform Fast Fourier Transform (FFT) on the data within the window to calculate gravity frequency, mean square frequency, frequency variance, band power, and relative energy.

[0037] Time-frequency domain features (3 items): Wavelet transform based on db4 wavelet basis to calculate wavelet entropy, marginal spectral peak and instantaneous frequency mean.

[0038] Fitting features (4 items): Perform linear / quadratic fitting on the parameters within the window, and calculate the first derivative, depth gradient, linear fitting slope, and quadratic fitting trend coefficient.

[0039] That is, for any base sample in multiple base sample groups within each depth sliding window (e.g., 50 drilling depths in 50 base sample groups within a short-term depth sliding window), the aforementioned 24 (m=24) statistical feature parameters are calculated, thus obtaining an extended sample group. After all depth sliding windows are calculated, multiple extended sample groups are obtained, each containing 9×24=216 statistical feature parameters. Finally, for three depth sliding windows at three different scales, three extended sample sets are generated, plus the base sample sets, resulting in a total of four sample datasets. Sample dataset 1: Basic sample set.

[0040] Sample dataset 2: Extended sample set corresponding to short-term depth sliding window.

[0041] Sample dataset 3: Extended sample set corresponding to the mid-term depth sliding window.

[0042] Sample dataset 4: Extended sample set corresponding to long-term depth sliding window.

[0043] Step S140: For each of the extended sample sets, based on the correlation between the statistical feature parameters, optimize the statistical feature parameters in the extended sample set to obtain an optimized first extended sample subset. Each extended sample group in the extended sample set includes multiple statistical feature parameters. When the number of statistical feature parameters is large (e.g., 2^16), although it can improve the accuracy of model training, it increases the computational load and time consumption of model training. Therefore, in this step, the correlation between statistical feature parameters is calculated, and only one of the two statistical feature parameters with high correlation is retained, thereby obtaining the optimized first extended sample subset. It can be understood that the number of extended sample groups in each first extended sample subset remains unchanged, only the number of statistical feature parameters in each extended sample group is reduced.

[0044] Step S150: Obtain the rock mass mechanical parameter labels corresponding to the basic sample set and each of the first extended sample subsets. Specifically, the true values ​​of rock mass mechanical parameters varying with depth obtained from core sampling after drilling is completed can be obtained, and the true values ​​at the corresponding depths can be used as the rock mass mechanical parameter labels corresponding to the sample groups of the basic sample set and each of the first extended sample subsets.

[0045] Step S160: For each of the first extended sample subsets, based on the importance of each statistical feature parameter to the rock mechanics parameter label, optimize the statistical feature parameters in each of the first extended sample subsets to obtain the optimized second extended sample subsets.

[0046] To further reduce the computational load during model training, this step focuses on the importance of each statistical feature parameter to the rock mechanics parameter labels. A higher importance statistical feature parameter indicates a greater impact on the rock mechanics parameters and also affects the prediction accuracy of the subsequent rock mechanics parameter determination model. Therefore, statistical feature parameters with higher importance to the rock mechanics parameter labels are retained. This reduces the number of statistical feature parameters in each extended sample group without affecting the prediction accuracy of the rock mechanics parameter determination model, thereby reducing the computational load of model training and improving the training efficiency of the model.

[0047] Step S170: Based on the basic sample set, each second extended sample subset and its corresponding rock mass mechanics parameter labels, train a rock mass mechanics parameter measurement model. The rock mass mechanics parameter measurement model is used to predict rock mass mechanics parameters based on the real-time collected basic sample set and its corresponding second extended sample subsets.

[0048] The rock mechanics parameter measurement model can use existing model structures as initial models for training. For rock mechanics parameters with continuous values, regression models can be used, such as Support Vector Machine (SVR) with Support Vector Regression, XGBoost (eXtreme Gradient Boosting) regression models, or BP (back propagation) neural network models. For rock mechanics parameters with Boolean values ​​(i.e., yes or no, or, 0 or 1), regression models can be used, such as Support Vector Machine (SVM) with Support Vector Classification, LightGBM (Lightweight Gradient Boosting) classification models, or random forest classification models.

[0049] During training, the basic sample set, each second extended sample subset, and their corresponding rock mass mechanics parameter labels are divided into a training set (60%), a test set (20%), and a validation set (20%) according to a certain ratio, such as 6:2:2. The rock mass mechanics parameter measurement models corresponding to different types of rock mass mechanics parameters are trained, tested, and validated, and finally the trained rock mass mechanics parameter measurement models are obtained.

[0050] The rock mass mechanics parameter measurement model constructed by the method in this embodiment is used to predict rock mass mechanics parameters based on the real-time collected basic sample set and the corresponding second extended sample subsets. It can realize the real-time measurement of rock mass mechanics parameters during drilling. Moreover, during the training process, features that characterize rock mass mechanics knowledge are introduced as input parameters of the rock mass mechanics parameter measurement model. Furthermore, statistical feature parameters related to the local microscopic properties, transitional properties, and overall macroscopic properties of the rock mass are added as input parameters through the depth sliding window mechanism to train the rock mass mechanics parameter measurement model. This allows the trained rock mass mechanics parameter measurement model to predict rock mass mechanics parameters more accurately.

[0051] In some embodiments, step S110, obtaining multiple sets of mechanical parameter sample groups during the drilling process, specifically includes: Step S111: Obtain multiple sets of original mechanical parameters during the drilling process. Specifically, multiple sets of original mechanical parameters can be obtained by directly acquiring the data collected by each sensor.

[0052] Step S112: Based on the drilling depth, drilling pressure, and drilling speed in each original mechanical parameter set, as well as the constraints of the stable drilling stage, determine whether each original mechanical parameter set corresponds to the stable drilling stage.

[0053] Because the drilling process follows a cyclical pattern of "drilling-pulling-preparation" (drilling being the stable drilling phase, while pulling and preparation are unstable drilling phases), the collected mechanical parameters will be mixed with data from unstable drilling phases such as pulling and preparation and drill pipe connection, thus affecting the accuracy of subsequent model training. Existing related technologies do not consider drill pipe length limitations (a typical drill pipe length is 3m, with an effective drilling length of approximately 2.7m, or 0.9×3m), and the drilling data is sorted by time rather than depth, making it impossible to match the depth distribution characteristics of rock mass parameters. Therefore, in this step, by using the drilling depth, drilling pressure, and drilling speed from the original mechanical parameter sets, as well as the constraints of the stable drilling phase, we determine whether each original mechanical parameter set corresponds to a stable drilling phase, thereby obtaining the original mechanical parameter set corresponding to the stable drilling phase.

[0054] Specifically, the constraints for the stable drilling phase include: A: Determination of the starting point of the stable drilling stage: (1).

[0055] B: Determination of the end point of the stable drilling stage: (2).

[0056] in, n This indicates the drilling speed (r / min). n min This indicates the minimum drilling speed, for example: 10% of the drilling rig's rated speed. P Indicates drilling pressure. P min This indicates the minimum drilling pressure (kPa), for example, 5% of the maximum thrust of the drilling rig.

[0057] C: Drilling depth constraint: (3).

[0058] in, H sThis indicates the rock mass depth corresponding to the start of the stable drilling stage. H e This indicates the rock mass depth corresponding to the end of the stable drilling stage. L lim This indicates the actual length of the drill pipe. This constraint ensures that the extracted steady-state phase data covers the effective drilling length of the drill pipe, avoiding depth gaps caused by not drilling a single drill pipe completely.

[0059] D: Drilling pressure fluctuation constraint: |Δ P | < 5% P avg (4).

[0060] in, P avg To stabilize the average pressure during the drilling phase, Δ P The pressure change within a time interval Δt = 0.5s is defined as the amount of pressure change. This constraint ensures that data is collected under stable drilling pressure conditions, avoiding interference from sudden pressure changes caused by equipment failure.

[0061] E: Drilling speed stability constraint: | n - n set |<2r / min (5)。

[0062] in, n set This indicates the set speed of the drilling rig. This constraint ensures that data is collected under stable drilling speed conditions, avoiding interference during the transition phase of speed adjustment.

[0063] Step S113: Determine the original mechanical parameter set corresponding to the stable drilling stage as the mechanical parameter sample set.

[0064] Based on the above constraints, using the effective drilling length of a single drill pipe as the basic unit, the data is stitched together in an ordered manner according to the following steps: For each stable drilling cycle, record its starting depth. H s With the endpoint depth H e For discrete data within each stable cycle, if there is data loss due to sensor failure or communication failure, linear interpolation is used to complete it into continuous data with a depth interval of 0.01m (i.e., each 0.01m depth corresponds to a complete set of mechanical parameter samples). The interpolated data of all stable cycles are spliced ​​together according to the rock mass depth from shallow to deep to form mechanical parameter sample sets corresponding to each drilling depth, drilling speed, drilling pressure, drill rod torque, drilling speed and drilling current in an orderly manner according to depth.

[0065] In this embodiment, only the original mechanical parameter set corresponding to the stable drilling stage is selected as the mechanical parameter sample set to avoid the influence of parameters in the unstable drilling stage on the model during subsequent model training.

[0066] In some embodiments, prior to step S120, the method further includes: Step S114: Use the Kalman filter algorithm to denoise the drilling speed in the mechanical parameter sample group.

[0067] Step S115: Use wavelet threshold denoising algorithm to denoise the drilling pressure and drill rod torque in the mechanical parameter sample group respectively.

[0068] Step S116: Use the adaptive noise complete set empirical mode decomposition algorithm to denoise the drilling current in the mechanical parameter sample group.

[0069] Specifically, to ensure the accuracy of the collected mechanical parameter sample data, denoising processing is required for certain mechanical parameter samples. Traditional denoising methods all use a single denoising algorithm to denoise all parameters, leading to distortion in some mechanical parameter samples after denoising. For example, spike noise in pressure signals cannot be removed, and depth signals are over-smoothed. Therefore, in this embodiment, by comparing the denoising effects of five mainstream denoising methods (moving average, wavelet thresholding, Kalman filtering, empirical mode decomposition (EMD), and adaptive noise ensemble empirical mode decomposition (CEEMDAN)) (using signal-to-noise ratio (SNR) and root mean square error (RMSE) as evaluation indicators), differentiated denoising strategies are matched for different mechanical parameter samples. The specific steps for selecting differentiated denoising strategies for different mechanical parameter samples are as follows: Step 1: Select three sets of drilling data for different geological conditions (e.g., sandstone formation, granite formation, fractured zone formation). Each set of data contains 1,000 sampling points, which is 1,000 sets of mechanical parameter samples.

[0070] Step 2: Drilling speed in each set of mechanical parameter samples v Drilling pressure P Drill pipe torque T and drilling current Cur Five mainstream denoising methods were used for processing, and the improvement in signal-to-noise ratio (SNR) and the reduction in root mean square error (RMSE) were used as evaluation indicators. The denoising results are shown in Table 1 below: Table 1. Comparison of denoising results from different denoising methods parameter Moving average Wavelet thresholding for noise reduction Kalman filtering EMD CEEMDAN Optimal denoising method SNR +10dB RMSE -18% SNR +11dB RMSE -22% SNR +15dB RMSE -30% SNR +12dB RMSE -25% SNR +13dB RMSE -28% Kalman filtering SNR +12dB RMSE -20% SNR +20dB RMSE -40% SNR +14dB RMSE -25% SNR +16dB RMSE -32% SNR +17dB RMSE -35% Wavelet thresholding for noise reduction SNR +11dB RMSE -19% SNR +19dB RMSE -38% SNR +13dB RMSE -24% SNR +15dB RMSE -30% SNR +16dB RMSE -33% Wavelet thresholding for noise reduction SNR +9dB RMSE -16% SNR +13dB RMSE -26% SNR +11dB RMSE -21% SNR +14dB RMSE -28% SNR +15dB RMSE -31% CEEMDAN Furthermore, the parameters of the selected optimal method are further optimized, such as selecting a "heuristic threshold" for wavelet threshold denoising and setting the noise intensity coefficient of CEEMDAN to 0.2 to ensure that the drilling data is sufficiently denoised.

[0071] In some embodiments, step S120, based on each set of mechanical parameter sample groups and rock mechanics theory, determines a set of reconstructed parameter sample groups characterizing the mechanical properties of the rock mass, specifically including: Step S121: Drilling pressure based on the mechanical parameter sample group P Drilling speed v Drill pipe torque T and drilling speed n Based on the aforementioned rock mechanics theory, the drilling specific energy, which characterizes the hardness of the rock mass, is determined.

[0072] Specifically, the drilling specific energy is calculated using the following formula. SE : (6).

[0073] in, R Where is the borehole radius.

[0074] Step S122: Based on the average values ​​of drilling pressure, drilling speed, and drilling rotation speed in the mechanical parameter sample group at the preset drilling displacement, and in conjunction with the rock mechanics theory, determine the rock mass strength correlation coefficient. α .

[0075] Specifically, the rock mass strength correlation coefficient is calculated using the following formula. α : (7).

[0076] in, A Indicates the borehole area. v avg , P avg and n avg These represent the average drilling speed, average drilling pressure, and average drilling rotation speed within each 0.5m drilling depth, respectively.

[0077] Step S123: Based on the drilling pressure, drilling speed, drilling rotation speed and the correlation coefficient of the rock mass strength in the mechanical parameter sample group, and in combination with the rock mechanics theory, determine the drilling power index that characterizes the drillability of the rock mass.

[0078] Specifically, the drilling power index is calculated using the following formula. DPI : (8).

[0079] Step S124: Form a set of reconstructed parameter sample groups based on the drilling specific energy, rock mass strength correlation coefficient and drilling power index.

[0080] It should be noted that the reconstructed parameter sample group is not limited to the drilling specific energy, rock mass strength correlation coefficient and drilling power index mentioned above, but can also be other reconstructed parameter samples that characterize the mechanical properties of rock mass.

[0081] In some embodiments, step S140, for each of the extended sample sets, optimizes the statistical feature parameters in the extended sample set based on the correlation between the statistical feature parameters to obtain an optimized first extended sample subset, specifically including: Step S141: For any of the extended sample sets, calculate the variance of each statistical feature parameter in each extended sample group, and delete statistical feature parameters whose variance is less than a preset variance threshold. The preset variance threshold can be set according to actual conditions; for example, the preset variance threshold can be set to 10. -3 The average drilling speed varies very little across different formations, less than 10. -3 Therefore, after deleting the average drilling speed and statistical feature parameters with variances less than the preset variance threshold, the retention rate of statistical feature parameters in each extended sample group is approximately 80%.

[0082] Step S142: Calculate the correlation coefficients of each pair of statistical feature parameters among the remaining statistical feature parameters. If the correlation coefficient is greater than a preset correlation coefficient threshold, retain one of the pair of statistical feature parameters to obtain the first expanded sample subset. For example, the Pearson correlation coefficient can be calculated, with a preset correlation coefficient threshold of 0.8. For instance, the correlation coefficient between the average drilling pressure and the drilling specific energy is 0.85, which is greater than 0.8, so only one of them needs to be retained. After filtering based on the correlation coefficient, the retention rate of statistical feature parameters in each expanded sample group is approximately 60%.

[0083] In this embodiment, by removing redundant statistical feature parameters, the dimensionality of each expanded sample group is reduced, shortening the model training time by approximately 40%. Simultaneously, it eliminates the risk of overfitting caused by feature redundancy, addressing the shortcomings of existing technologies such as low efficiency and susceptibility to overfitting during full-feature training. For example, for the aforementioned sample datasets 2-4, the feature dimensions of the expanded sample groups in their respective optimized first expanded sample subsets are reduced from 216 dimensions to 130, 96, and 80 dimensions, respectively.

[0084] In some embodiments, step S150, obtaining the rock mass mechanical parameter labels corresponding to the basic sample set and each first extended sample subset, specifically includes: Step S151: Obtain the true values ​​of rock mass mechanical parameters that vary with depth obtained from core sampling after drilling is completed. For example, the true values ​​of rock mass mechanical parameters may include the true values ​​of uniaxial compressive strength (UCS), elastic modulus (E), and fault fracture zone distribution.

[0085] Step S152: Interpolate the true values ​​of the rock mass mechanical parameters according to a preset minimum depth step to obtain the rock mass mechanical parameter labels corresponding to each basic sample group in the basic sample set. The mechanical parameter sample groups in each basic sample group are obtained based on the preset minimum depth step, which is, for example, the aforementioned depth interval of 0.01m.

[0086] For rock mechanics parameters with continuous values ​​(such as UCS and E), Kriging interpolation can be used to fill in the missing values ​​within the depth interval.

[0087] For rock mechanics parameters with Boolean values ​​(e.g., whether it is a fault fracture zone), the nearest depth voting interpolation method can be used. For the missing depth H, take the depth points with three known labels above and below it (H-3m, H-2m, H-1m, H+1m, H+2m, and H+3m), and count the number of labels with 1 (fault fracture zone). If the number is ≥4, fill in 1; otherwise, fill in 0 (non-fault fracture zone).

[0088] Step S153: After interpolation, the rock mechanics parameter labels are associated with the base sample group in the base sample set and the extended sample group in each extended sample set according to the depth. After association, a dataset with both existing samples and corresponding labels is formed, which provides sufficient samples for subsequent model pre-training and training, and solves the problems of missing label data and insufficient model training samples in the existing technology.

[0089] In some embodiments, step S160: For each first extended sample subset, based on the importance of each statistical feature parameter to the rock mechanics parameter label, optimize the statistical feature parameters in each first extended sample subset to obtain optimized second extended sample subsets, including: Step S161: For any first extended sample subset, based on each extended sample group and the corresponding rock mechanics parameter label in the first extended sample subset, train and validate the pre-trained model to obtain the feature importance score of each statistical feature parameter output by the pre-trained model. The feature importance score characterizes the importance of the corresponding statistical feature parameter to the rock mechanics parameter label.

[0090] Specifically, any first extended sample subset and its corresponding labels are divided into a pre-training set (80%) and a validation set (20%) in a ratio of 8:2, and then a dedicated pre-training model is selected according to the value type of the rock mass mechanics parameters.

[0091] For rock mechanics parameters with continuous values ​​(such as UCS and E), a machine learning regression model for feature importance evaluation, such as the random forest (based on CART decision tree ensemble) model, is used to output the importance score of each statistical feature parameter in any first extended sample subset, taking advantage of its anti-overfitting properties and strong adaptability to numerical data.

[0092] For rock mechanics parameters with Boolean values ​​(e.g., whether it is a fault fracture zone), a random forest classification model (based on Gini coefficients to divide decision nodes) is used. Taking advantage of its high accuracy in classifying discrete labels, the model outputs the feature importance score of each statistical feature parameter.

[0093] Whether it's a regression-based random forest model or a classification-based random forest model, the feature importance scores of each statistical feature parameter can be obtained using the Python function rf.feature_importances_. This function normalizes the feature importance scores of each statistical feature parameter to the range [0,1], and the sum of the feature importance scores of each statistical feature parameter is 1.

[0094] During pre-training, the structural parameters of the machine learning algorithm for feature importance evaluation are determined through K-fold cross-validation and grid search, and the regression model uses Ri... 2 The classification model is determined by accuracy, and the F1-Score is used when the samples are imbalanced, thus obtaining the pre-trained model with the highest accuracy.

[0095] Regardless of whether it's a regression or classification model, taking random forest as an example of a pre-trained model, the model parameters and optimization target metrics during model pre-training are shown in Table 2 below: Table 2 Pre-trained model parameters and optimization metrics Model type Parameters to be optimized Parameter search range Optimize target indicators Random Forest Regression Number of trees (n_estimators), maximum depth (max_depth), minimum number of sample splits (min_samples_split) n_estimators:50-200; max_depth:3-15; min_samples_split:2-10 <![CDATA[Validation set R 2 ≥0.80]]> Random Forest Classification Number of trees, maximum number of features (max_features), minimum number of leaf nodes (min_samples_leaf) n_estimators:50-200; max_features: "sqrt" / "log2"; min_samples_leaf:1-5 Validation set accuracy ≥ 90% (when data is balanced), F1-Score ≥ 0.85 (when data is imbalanced) Step S162: Sort the feature importance scores from largest to smallest, and accumulate the feature importance scores according to the sorting. When the accumulated score exceeds the score threshold, stop the accumulation and retain the statistical feature parameters that participated in the accumulation in each of the extended sample groups of any first extended sample subset to obtain the second extended sample subset corresponding to any first extended sample subset.

[0096] The scoring threshold is set according to the actual situation. Taking the feature importance score of each statistical feature parameter obtained by the function rf.feature_importances_ as an example, the scoring threshold can be 90%. Taking the above sample datasets 2 to 4 as examples, the feature dimensions of the extended sample groups in the first extended sample subsets after optimization are reduced from 216 dimensions to 130, 96, and 80 dimensions, respectively. In this step, the first extended sample subset corresponding to sample dataset 2 has a total of 130 statistical feature parameters. If the feature importance score of the first 25 statistical feature parameters is greater than 90%, then the first 25 statistical feature parameters are retained to obtain the second extended sample subset corresponding to sample dataset 2. The same operation is performed for sample datasets 3 and 4. Finally, each extended sample group in sample dataset 2 is reduced from 130 dimensions to about 25 dimensions; each extended sample group in sample dataset 3 is reduced from 96 dimensions to about 13 dimensions; and each extended sample group in sample dataset 4 is reduced from 80 dimensions to about 16 dimensions.

[0097] In this embodiment, the interference of irrelevant features on the prediction results is eliminated by using the feature importance scores of each statistical feature parameter output by the pre-trained model, further reducing the feature dimension in the expanded sample group, thereby improving the training efficiency of the final rock mass mechanics parameter determination model by 20%.

[0098] In some embodiments, step S170, based on the basic sample set, each second extended sample subset, and their respective corresponding rock mechanics parameter labels, trains a rock mechanics parameter measurement model, specifically including: Based on the basic sample set, each second extended sample subset, and their corresponding rock mechanics parameter labels, multiple candidate models are trained, tested, and validated, and the candidate model with the best model stability index is selected as the rock mechanics parameter determination model.

[0099] Specifically, the basic sample set, each second extended sample subset, and their corresponding rock mechanics parameter labels are divided into a training set (60%), a test set (20%), and a validation set (20%) in a 6:2:2 ratio. For rock mechanics parameters with continuous values, a regression model can be selected; for rock mechanics parameters with Boolean values, a classification model can be selected. Multiple regression and classification models can be selected for training, testing, and validation. The regression model and classification model with the highest prediction accuracy are then used as the rock mechanics parameter determination models. For example, the selection of regression and classification models and the optimal model parameter range are shown in Table 3 below.

[0100] Table 3. Selection of Regression and Classification Models and Optimal Model Parameter Range Target parameter type Candidate machine learning algorithms Optimization algorithm (optimizing model parameters) Optimization parameter range Optimal Algorithm Example Continuous type (e.g., UCS) Support Vector Regression (SVR), XGBoost Regression, Backpropagation Neural Network Particle Swarm Optimization (PSO) SVR: Penalty coefficient C∈[0.1,10], kernel function parameter γ∈[0.01,1]; XGBoost: Learning rate η∈[0.01,0.3], tree depth max_depth∈[3,10] XGBoost regression (η=0.1, max_depth=6, RMSE=3.2%) Boolean type (e.g., fault fracture zone) Support Vector Classification (SVC), LightGBM Classification, Random Forest Classification Genetic Algorithm (GA) SVC: Penalty coefficient C∈[0.1,10], kernel function parameter γ∈[0.01,1]; LightGBM: Learning rate η∈[0.01,0.3], number of leaf nodes num_leaves∈[31,127] LightGBM classification (η=0.08, num_leaves=63, F1-Score=0.91) The optimal parameters obtained through optimization are tested on the validation set, requiring that the deviation of the evaluation metrics between the validation set and the test set be less than 3% (e.g., test set R). 2 =0.88, validation set R 2 To ensure model stability, the model group with the smallest deviation of the evaluation index between the test set and the validation set is ultimately selected as the rock mechanics parameter determination model.

[0101] It is understandable that: for all rock mechanics parameters with continuous values, the same regression model can be used as the initial model for training the rock mechanics parameter measurement model; for all rock mechanics parameters with Boolean values, the same classification model can be used as the initial model for training the rock mechanics parameter measurement model, ultimately resulting in two rock mechanics parameter measurement models.

[0102] This invention also provides a method for determining rock mass mechanical parameters while drilling, such as... Figure 2 As shown, it includes the following steps S210 to S230.

[0103] Step S210: Obtain the basic parameter set and the corresponding second extended parameter subsets for real-time acquisition. Specifically, refer to steps S110 to S120 above to obtain the basic parameter set, and follow step S130 above to obtain the extended parameter set. For the same depth sliding window, retain the statistical feature parameters corresponding to the second extended sample subset in the corresponding extended parameter subset, thus obtaining the second extended parameter subset.

[0104] Step S220: Input the basic parameter set and the corresponding second extended parameter subsets into the rock mechanics parameter measurement model constructed by the rock mechanics parameter measurement model construction method of any of the above embodiments, and obtain the candidate rock mechanics parameters corresponding to the basic parameter set and each second extended parameter subset output by the rock mechanics parameter measurement model.

[0105] Step S230: Determine the final rock mass mechanics parameters based on the candidate rock mass mechanics parameters corresponding to the basic parameter set and each of the second extended parameter subsets. Since the basic parameter set and each of the second extended parameter subsets will predict the corresponding rock mass mechanics parameters when input into the rock mass mechanics parameter determination model, it is necessary to determine the final rock mass mechanics parameters.

[0106] For example, the rock mass mechanics parameter measurement model itself has high prediction accuracy. For rock mass mechanics parameters with continuous values, the average value of the predicted candidate rock mass mechanics parameters can be taken as the final rock mass mechanics parameter. For rock mass mechanics parameters with Boolean values, the average value of the predicted candidate rock mass mechanics parameters can also be taken. If the average value is greater than or equal to 0.5, it is taken as 1; otherwise, it is taken as 0.

[0107] The rock mechanics parameter determination method in this embodiment uses the rock mechanics parameter determination model constructed by the rock mechanics parameter determination model construction method of any of the above embodiments, and predicts the rock mechanics parameters based on the basic parameter set collected during drilling and the corresponding second extended parameter subsets, which can predict the rock mechanics parameters in real time and accurately.

[0108] In step S230 above, the final rock mass mechanical parameters are obtained by averaging the candidate rock mass mechanical parameters. However, this does not consider the difference in prediction accuracy of the rock mass mechanical parameter determination model for different datasets. Low-precision candidate rock mass mechanical parameters reduce the accuracy of the final rock mass mechanical parameters. Therefore, in some embodiments, step S230, based on the candidate rock mass mechanical parameters corresponding to the basic parameter set and each second extended parameter subset, determines the final rock mass mechanical parameters, specifically including: Step S231: For rock mechanics parameters with continuous values, determine the deviation based on the maximum and minimum candidate rock mechanics parameters. If the deviation is less than or equal to a preset deviation threshold, obtain the final rock mechanics parameters by weighting each candidate rock mechanics parameter and its corresponding model test accuracy. Otherwise, delete the candidate rock mechanics parameter with the largest deviation, and obtain the final rock mechanics parameters by weighting the remaining candidate rock mechanics parameters and their corresponding model test accuracy. The model test accuracy is the accuracy value of the basic sample set and each second extended sample subset during the testing phase of model training. The deviation is the difference between each candidate rock mechanics parameter value and the mean value of each candidate rock mechanics parameter value; the larger the difference, the greater the deviation.

[0109] Specifically, the formula for determining the deviation is as follows: (Y max -Y min ) / Y min (9).

[0110] Among them, Y max Y represents the maximum value of the candidate rock mass mechanical parameters. min This represents the minimum value of the candidate rock mass mechanical parameters. The preset deviation threshold can be set according to the actual situation, for example: 20%.

[0111] The weighted formula for the candidate rock mass mechanical parameters and their corresponding model test accuracy is as follows: (10).

[0112] in, Y final This represents the final rock mass mechanics parameters. Y i Indicates the first i Candidate rock mass mechanical parameters. Indicates the firsti Model test accuracy corresponding to each sample set k This indicates the number of sample sets (including the basic sample set and the corresponding second extended sample subsets).

[0113] For example, taking the four sample datasets and the predicted UCSs mentioned above, the four sample datasets correspond to four predicted UCSs: UCS1, UCS2, UCS3, and UCS4. (UCS) max - UCS min ) / UCS min In the case of ≤20%, the UCS is calculated directly according to formula (10). final At this point, k=4.

[0114] For example: UCS1 = 80MPa corresponds to sample dataset 1. R 2 =0.83, UCS2 corresponding to sample dataset 2 is 82MPa. R 2 =0.82, UCS3=72MPa for sample dataset 3. R 2 =0.89, UCS4=84MPa for sample dataset 4. R 2 =0.68, calculate (UCS) max - UCS min ) / UCS min =(84-72) / 72≈16.7%≤20%, which satisfies the direct weighting condition, and the final UCS final = 79.1MPa.

[0115] In (UCS) max - UCS min ) / UCS min In cases where the deviation is greater than 20%, the UCS with the largest deviation is deleted. For example, if the predicted UCS3 of sample dataset 3 is 60 MPa, and the predicted values ​​of the other three model datasets are between 80 and 84 MPa, the UCS3 corresponding to sample dataset 3 is deleted. Then, the remaining UCSs are substituted into formula (10) to calculate the UCS. final At this point, k=3.

[0116] Step S232: For rock mechanics parameters with Boolean values, the final rock mechanics parameter is determined from the candidate rock mechanics parameters based on the majority voting method. When the number of votes for each candidate rock mechanics parameter with different values ​​is equal, a weighted calculation is performed using the predicted probability of each candidate rock mechanics parameter with different values ​​and the corresponding F1 score as weights to obtain the comprehensive probability of different values. The value with the larger comprehensive probability is selected as the final value of the rock mechanics parameter. The F1 score is obtained by testing the basic sample set and each second extended sample subset during the testing phase of the model training process. Each sample set corresponds one-to-one with each parameter set, and each sample set corresponds to one F1 score.

[0117] Specifically, for the majority voting method, for example, in the candidate rock mass mechanical parameters of four sample datasets (0 = non-fractured zone, 1 = fractured zone), if a certain value appears ≥3 times (e.g., the prediction is 1 based on three sample datasets and 0 based on one sample dataset), then the final rock mass mechanical parameter is 1.

[0118] When the number of votes for candidate rock mass mechanical parameters with different values ​​is equal, the overall probability is calculated using the following formula: (11); (12).

[0119] in, P (0) and P (1) represent the combined probability respectively. P i (0) and P i (1) respectively represent the first i The probability that the candidate rock mass mechanics parameters corresponding to each parameter set (the basic parameter set and each second extended parameter subset) are 0 and 1. F 1 i Indicates the first i The F1 score obtained by the sample set corresponding to each parameter set during the testing phase of the model training process.

[0120] In some embodiments, during the application stage of the rock mass mechanics parameter determination model, after the rock mass mechanics parameters are inferred, they are compared and verified with the measured rock mass parameters obtained from borehole core sampling. The verification indicators are as follows: Continuous parameters: absolute error ≤10%; Label-type parameters: Classification accuracy ≥ 80% (under complex geological conditions), classification accuracy ≥ 90% (under simple geological conditions).

[0121] Meanwhile, after drilling is completed, a borehole television observation instrument is used to obtain rock mass structure images (such as the location of fracture zones and the degree of fracture development), and compares them with the depth corresponding to the model prediction results. The depth matching deviation is required to be ≤0.2m (e.g., the model predicts that the fracture zone is at 25.3m, but the actual observation shows that the fracture zone is at 25.45m).

[0122] If the above verification indicators fail to meet the standards, new geological conditions emerge (such as transitioning from sandstone strata to granite strata), or a total of 5 new boreholes are completed or the cumulative drilling length exceeds 100m, the rock mass mechanics parameter measurement model will be updated.

[0123] The update strategy is as follows: Sensor drift: If the verification finds that the acquisition error of a certain parameter is >3% (such as zero drift of the pressure sensor), recalibrate the sensor and collect 10 sets of standard working condition data.

[0124] Feature selection bias: Adjust the cumulative threshold of feature importance (e.g., from 90% to 80% to retain more potential features).

[0125] Changes in geological conditions: Add the "drilling data + measured parameters" from the new borehole to the database (ensuring that the proportion of new samples is ≥10%), and re-execute steps S110 to S170 above to retrain the model in order to update the model parameters and adapt the rock mechanics parameter measurement model to the new geological conditions.

[0126] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include: a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communications interface 320, and the memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute a method for constructing a rock mass mechanics parameter determination model, which includes: Multiple sets of mechanical parameter samples are obtained during the drilling process, and each set of mechanical parameter samples includes multiple mechanical parameter samples.

[0127] Based on each set of mechanical parameter samples and rock mechanics theory, a set of reconstructed parameter samples characterizing the mechanical properties of rock mass is determined to obtain a basic sample set consisting of multiple sets of mechanical parameter samples and multiple sets of reconstructed parameter samples.

[0128] Based on multiple depth sliding windows of different scales, m statistical feature parameters of each of the n basic samples in each depth sliding window of each scale are calculated to obtain multiple extended sample sets. Each extended sample set includes multiple extended sample groups, and each extended sample group includes n×m statistical feature parameters, where n and m are both greater than or equal to 1.

[0129] For each of the extended sample sets, the statistical feature parameters in the extended sample set are optimized based on the correlation between the statistical feature parameters to obtain the optimized first extended sample subset.

[0130] Obtain the rock mechanics parameter labels corresponding to the basic sample set and each first extended sample subset.

[0131] For each of the first extended sample subsets, the statistical feature parameters in each of the first extended sample subsets are optimized based on the importance of each statistical feature parameter to the rock mechanics parameter label, resulting in optimized second extended sample subsets.

[0132] Based on the basic sample set, each second extended sample subset, and their corresponding rock mass mechanics parameter labels, a rock mass mechanics parameter measurement model is trained. The rock mass mechanics parameter measurement model is used to predict rock mass mechanics parameters based on the real-time collected basic sample set and the corresponding second extended sample subsets.

[0133] Alternatively, a method for determining rock mechanics parameters while drilling may be employed, which includes: Obtain the basic parameter set for real-time acquisition and the corresponding subsets of second extended parameters.

[0134] The basic parameter set and the corresponding second extended parameter subsets are respectively input into the rock mass mechanics parameter determination model constructed by the rock mass mechanics parameter determination model construction method of any of the above embodiments, so as to obtain the candidate rock mass mechanics parameters corresponding to the basic parameter set and each second extended parameter subset output by the rock mass mechanics parameter determination model.

[0135] Based on the basic parameter set and the candidate rock mechanics parameters corresponding to each of the second extended parameter subsets, the final rock mechanics parameters are determined.

[0136] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0137] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is able to execute the rock mass mechanics parameter determination model construction method provided by the above methods, the method comprising: Multiple sets of mechanical parameter samples are obtained during the drilling process, and each set of mechanical parameter samples includes multiple mechanical parameter samples.

[0138] Based on each set of mechanical parameter samples and rock mechanics theory, a set of reconstructed parameter samples characterizing the mechanical properties of rock mass is determined to obtain a basic sample set consisting of multiple sets of mechanical parameter samples and multiple sets of reconstructed parameter samples.

[0139] Based on multiple depth sliding windows of different scales, m statistical feature parameters of each of the n basic samples in each depth sliding window of each scale are calculated to obtain multiple extended sample sets. Each extended sample set includes multiple extended sample groups, and each extended sample group includes n×m statistical feature parameters, where n and m are both greater than or equal to 1.

[0140] For each of the extended sample sets, the statistical feature parameters in the extended sample set are optimized based on the correlation between the statistical feature parameters to obtain the optimized first extended sample subset.

[0141] Obtain the rock mechanics parameter labels corresponding to the basic sample set and each first extended sample subset.

[0142] For each of the first extended sample subsets, the statistical feature parameters in each of the first extended sample subsets are optimized based on the importance of each statistical feature parameter to the rock mechanics parameter label, resulting in optimized second extended sample subsets.

[0143] Based on the basic sample set, each second extended sample subset, and their corresponding rock mass mechanics parameter labels, a rock mass mechanics parameter measurement model is trained. The rock mass mechanics parameter measurement model is used to predict rock mass mechanics parameters based on the real-time collected basic sample set and the corresponding second extended sample subsets.

[0144] Alternatively, a method for determining rock mechanics parameters while drilling may be employed, which includes: Obtain the basic parameter set for real-time acquisition and the corresponding subsets of second extended parameters.

[0145] The basic parameter set and the corresponding second extended parameter subsets are respectively input into the rock mass mechanics parameter determination model constructed by the rock mass mechanics parameter determination model construction method of any of the above embodiments, so as to obtain the candidate rock mass mechanics parameters corresponding to the basic parameter set and each second extended parameter subset output by the rock mass mechanics parameter determination model.

[0146] Based on the basic parameter set and the candidate rock mechanics parameters corresponding to each of the second extended parameter subsets, the final rock mechanics parameters are determined.

[0147] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for constructing a rock mass mechanics parameter determination model provided by the methods described above, the method comprising: Multiple sets of mechanical parameter samples are obtained during the drilling process, and each set of mechanical parameter samples includes multiple mechanical parameter samples.

[0148] Based on each set of mechanical parameter samples and rock mechanics theory, a set of reconstructed parameter samples characterizing the mechanical properties of rock mass is determined to obtain a basic sample set consisting of multiple sets of mechanical parameter samples and multiple sets of reconstructed parameter samples.

[0149] Based on multiple depth sliding windows of different scales, m statistical feature parameters of each of the n basic samples in each depth sliding window of each scale are calculated to obtain multiple extended sample sets. Each extended sample set includes multiple extended sample groups, and each extended sample group includes n×m statistical feature parameters, where n and m are both greater than or equal to 1.

[0150] For each of the extended sample sets, the statistical feature parameters in the extended sample set are optimized based on the correlation between the statistical feature parameters to obtain the optimized first extended sample subset.

[0151] Obtain the rock mechanics parameter labels corresponding to the basic sample set and each first extended sample subset.

[0152] For each of the first extended sample subsets, the statistical feature parameters in each of the first extended sample subsets are optimized based on the importance of each statistical feature parameter to the rock mechanics parameter label, resulting in optimized second extended sample subsets.

[0153] Based on the basic sample set, each second extended sample subset, and their corresponding rock mass mechanics parameter labels, a rock mass mechanics parameter measurement model is trained. The rock mass mechanics parameter measurement model is used to predict rock mass mechanics parameters based on the real-time collected basic sample set and the corresponding second extended sample subsets.

[0154] Alternatively, a method for determining rock mechanics parameters while drilling may be employed, which includes: Obtain the basic parameter set for real-time acquisition and the corresponding subsets of second extended parameters.

[0155] The basic parameter set and the corresponding second extended parameter subsets are respectively input into the rock mass mechanics parameter determination model constructed by the rock mass mechanics parameter determination model construction method of any of the above embodiments, so as to obtain the candidate rock mass mechanics parameters corresponding to the basic parameter set and each second extended parameter subset output by the rock mass mechanics parameter determination model.

[0156] Based on the basic parameter set and the candidate rock mechanics parameters corresponding to each of the second extended parameter subsets, the final rock mechanics parameters are determined.

[0157] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0158] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0159] All actions involving the acquisition of signal information or data in this invention are carried out in compliance with the relevant data protection laws and policies of the country where the device is located, and with the authorization granted by the owner of the device.

[0160] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a model for determining rock mass mechanical parameters, characterized in that, include: Multiple sets of mechanical parameter samples are obtained during the drilling process, and each set of mechanical parameter samples includes multiple mechanical parameter samples; Based on each set of mechanical parameter samples and rock mechanics theory, a set of reconstructed parameter samples characterizing the mechanical properties of rock mass is determined to obtain a basic sample set consisting of multiple sets of mechanical parameter samples and multiple sets of reconstructed parameter samples. Based on multiple depth sliding windows of different scales, m statistical feature parameters of each of the n basic samples in each depth sliding window of each scale are calculated to obtain multiple extended sample sets. Each extended sample set includes multiple extended sample groups, and each extended sample group includes n×m statistical feature parameters, where n and m are both greater than or equal to 1. For each of the extended sample sets, the statistical feature parameters in the extended sample set are optimized based on the correlation between each statistical feature parameter to obtain the optimized first extended sample subset; Obtain the rock mass mechanics parameter labels corresponding to the basic sample set and each first extended sample subset; For each of the first extended sample subsets, based on the importance of each statistical feature parameter to the rock mechanics parameter label, the statistical feature parameters in each of the first extended sample subsets are optimized to obtain each of the optimized second extended sample subsets; Based on the basic sample set, each second extended sample subset, and their corresponding rock mass mechanics parameter labels, a rock mass mechanics parameter measurement model is trained. The rock mass mechanics parameter measurement model is used to predict rock mass mechanics parameters based on the real-time collected basic sample set and the corresponding second extended sample subsets.

2. The method for constructing a rock mass mechanical parameter determination model according to claim 1, characterized in that, Obtain multiple sets of mechanical parameter samples during the drilling process, including: Obtain multiple sets of original mechanical parameters during the drilling process; Based on the drilling depth, drilling pressure, and drilling speed in each original mechanical parameter set, as well as the constraints of the stable drilling stage, determine whether each original mechanical parameter set corresponds to the stable drilling stage. The original set of mechanical parameters corresponding to the stable drilling stage is determined as the mechanical parameter sample set.

3. The method for constructing a rock mass mechanical parameter determination model according to claim 1, characterized in that, Before determining the reconstructed parameter sample set characterizing the rock mass mechanical properties based on each set of mechanical parameter samples and rock mechanics theory, the following steps are also included: The drilling speed in the mechanical parameter sample group is denoised using a Kalman filter algorithm. The drilling pressure and drill rod torque in the mechanical parameter sample group were denoised using a wavelet threshold denoising algorithm. The drilling current in the mechanical parameter sample group is denoised using an adaptive noise complete set empirical mode decomposition algorithm.

4. The method for constructing a rock mass mechanical parameter determination model according to claim 1, characterized in that, Based on each set of mechanical parameter samples and rock mechanics theory, a set of reconstructed parameter samples characterizing the mechanical properties of rock mass is determined, including: Based on the drilling pressure, drilling speed, drill rod torque and drilling speed in the mechanical parameter sample group, and combined with the rock mechanics theory, the drilling specific energy characterizing the hardness of the rock mass is determined. Based on the average values ​​of drilling pressure, drilling speed, and drilling rotation speed in the mechanical parameter sample group at the preset drilling displacement, and combined with the rock mechanics theory, the rock mass strength correlation coefficient is determined. Based on the drilling pressure, drilling speed, drilling rotation speed and the correlation coefficient of rock mass strength in the mechanical parameter sample group, and combined with the rock mechanics theory, the drilling power index characterizing the drillability of the rock mass is determined. A set of reconstructed parameter samples is formed using the drilling specific energy, rock mass strength correlation coefficient, and drilling power index.

5. The method for constructing a rock mass mechanical parameter determination model according to claim 1, characterized in that, For each of the extended sample sets, based on the correlation between the statistical feature parameters, the statistical feature parameters in the extended sample set are optimized to obtain an optimized first extended sample subset, including: For any of the extended sample sets, calculate the variance of each statistical feature parameter in each extended sample group, and delete statistical feature parameters whose variance is less than a preset variance threshold. Calculate the correlation coefficient between each pair of statistical feature parameters among the remaining statistical feature parameters. If the correlation coefficient is greater than the preset correlation coefficient threshold, retain one of the pair of statistical feature parameters to obtain the first extended sample subset.

6. The method for constructing a rock mass mechanical parameter determination model according to claim 1, characterized in that, Obtaining the rock mass mechanics parameter labels corresponding to the basic sample set and each first extended sample subset, including: Obtain the true values ​​of rock mass mechanical parameters as a function of depth obtained from core sampling after drilling is completed; The true values ​​of the rock mass mechanical parameters are interpolated according to a preset minimum depth step to obtain the rock mass mechanical parameter labels corresponding to each basic sample group in the basic sample set. The mechanical parameter sample groups in each basic sample group are obtained based on the preset minimum depth step. After interpolation, the corresponding rock mechanics parameter labels are associated with the base sample group in the base sample set and the extended sample group in each extended sample set according to depth.

7. The method for constructing a rock mass mechanical parameter determination model according to claim 1, characterized in that, For each of the first extended sample subsets, based on the importance of each statistical feature parameter to the rock mechanics parameter labels, the statistical feature parameters in each of the first extended sample subsets are optimized to obtain the optimized second extended sample subsets, including: For any first extended sample subset, based on each extended sample group and the corresponding rock mechanics parameter label in the first extended sample subset, a pre-trained model is trained and validated to obtain the feature importance score of each statistical feature parameter output by the pre-trained model. The features are sorted from largest to smallest according to their importance scores, and the feature importance scores are accumulated according to the sorting. When the accumulated score exceeds the score threshold, the accumulation stops, and the statistical feature parameters that participated in the accumulation in each of the expanded sample groups of any first expanded sample subset are retained to obtain the second expanded sample subset corresponding to any first expanded sample subset.

8. The method for constructing a rock mass mechanical parameter determination model according to any one of claims 1 to 7, characterized in that, Based on the aforementioned basic sample set, each second extended sample subset, and their corresponding rock mass mechanics parameter labels, a rock mass mechanics parameter measurement model is trained, including: Based on the basic sample set, each second extended sample subset, and their corresponding rock mechanics parameter labels, multiple candidate models are trained, tested, and validated, and the candidate model with the best model stability index is selected as the rock mechanics parameter determination model.

9. A method for determining rock mass mechanical parameters while drilling, characterized in that, include: Obtain the real-time acquisition of the basic parameter set and the corresponding subsets of the second extended parameters; The basic parameter set and the corresponding second extended parameter subsets are respectively input into the rock mass mechanics parameter determination model constructed by the rock mass mechanics parameter determination model construction method as described in any one of claims 1 to 8, so as to obtain the candidate rock mass mechanics parameters corresponding to the basic parameter set and each second extended parameter subset output by the rock mass mechanics parameter determination model. Based on the basic parameter set and the candidate rock mechanics parameters corresponding to each of the second extended parameter subsets, the final rock mechanics parameters are determined.

10. The method for determining rock mass mechanical parameters while drilling according to claim 9, characterized in that, Based on the basic parameter set and the candidate rock mechanics parameters corresponding to each of the second extended parameter subsets, the final rock mechanics parameters are determined, including: For rock mechanics parameters with continuous values, the deviation is determined based on the maximum and minimum candidate rock mechanics parameters. If the deviation is less than or equal to a preset deviation threshold, the final rock mechanics parameters are obtained by weighting each candidate rock mechanics parameter and its corresponding model test accuracy. Otherwise, after deleting the candidate rock mechanics parameter with the largest deviation, the final rock mechanics parameters are obtained by weighting the remaining candidate rock mechanics parameters and their corresponding model test accuracy. Here, the model test accuracy is the accuracy value of the basic sample set and each second extended sample subset during the test phase of model training. For rock mechanics parameters with Boolean values, the final rock mechanics parameter is determined from the candidate rock mechanics parameters based on the majority voting method. When the number of votes for each candidate rock mechanics parameter with different values ​​is equal, a weighted calculation is performed using the predicted probability of each candidate rock mechanics parameter with different values ​​and the corresponding F1 score as weights to obtain the comprehensive probability of different values. The value with the larger comprehensive probability is selected as the final value of the rock mechanics parameter. The F1 score is obtained by testing the basic sample set and each second extended sample subset during the testing phase of the model training process. Each sample set corresponds one-to-one with each parameter set, and each sample set corresponds to one F1 score.

Images