A rock integrity prediction method based on drilling and cusum algorithm

Abstract

The application discloses a rock integrity prediction method based on drilling and a CUSUM algorithm, and relates to the technical field of coal mining. The application comprises the following steps: a drilling data acquisition device is installed on an advanced drilling machine, and the collected data is processed; a time window detection function is added to the CUSUM algorithm, an initial drift threshold is set, and it is monitored whether deviation occurs; when the cumulative sum exceeds the threshold and lasts for a preset time, it is determined that cracks exist, and the initial deviation point is marked as the crack position; crack density and rock mass quality indexes are calculated according to the detected crack condition, and a rock mass quality evaluation report is generated.

Description

Technical Field

[0001] This invention belongs to the field of coal mining technology, and in particular relates to a rock integrity prediction method based on drilling and CUSUM algorithms. Through real-time data transmission, analysis and prediction between the surface and underground, this method can provide a reference for TBM intelligent decision-making, effectively prevent safety accidents such as roadway collapse and machine jamming, and eliminates the need for manual recording on site, thus reducing manual underground operations. Background Technology

[0002] Traditional methods for predicting rock mass integrity often employ advanced drilling and coring, as well as ground-penetrating radar (GPR) detection. Advanced drilling requires manual operation to obtain rock cores, which are then assessed for integrity using the rock quality index (RQD). This process necessitates interrupting TBM (Tunnel Boring Machine) operations, resulting in low efficiency (single-hole drilling typically takes over 2 hours) and the risk of core breakage during the process, leading to RQD calculation errors. While GPR offers non-contact detection, strong electromagnetic interference downhole and variations in rock moisture content severely impact radar signal quality, resulting in a detection rate of less than 30% for micro-fractures (width < 1 mm). Furthermore, its detection depth is typically limited to 5-8 m ahead of the tunnel face, failing to meet the requirements for long-distance advanced early warning. Both methods also require manual data recording and analysis, leading to significant human error and delayed feedback (4-6 hours from detection to report generation), thus hindering real-time TBM decision-making.

[0003] To improve the timeliness of monitoring, some projects have introduced monitoring-while-drilling (MWD) technology based on onboard sensors. This technology indirectly inverts rock mass integrity by collecting parameters such as thrust, torque, drilling speed, rotational speed, and vibration. However, existing technologies have the following core shortcomings: (1) The data dimension is limited and the information utilization rate is low: it only relies on 5 mechanical and kinematic parameters and does not consider the microscopic signals (such as acoustic emission) and intuitive morphology (such as images of fractures in the borehole) when the rock mass is fractured. For example, accidental vibration interference in homogeneous strata is easily misjudged as fracture signals, and the small mechanical parameter changes in the early stage of fracture development may be masked by noise, resulting in a high false alarm rate (>25%) and false negative rate (>18%).

[0004] (2) Poor algorithm adaptability and insufficient detection accuracy: Existing technologies mostly use fixed threshold method or traditional CUSUM (cumulative control) algorithm for anomaly detection. Fixed threshold method cannot adapt to geological conditions with different lithologies (such as the difference of mechanical parameter benchmark values ​​between sandstone and mudstone can be up to 40%), and is prone to the problem of "different judgments for the same threshold"; traditional CUSUM algorithm lacks time window constraints, is sensitive to instantaneous noise, and has difficulty in identifying "small and continuous" fracture characteristic parameter shifts (such as thrust slowly increasing by 5%~8%), resulting in a fracture initiation position positioning error of more than 0.5m, which cannot accurately guide support operations.

[0005] (3) Disconnect between results feedback and application: The existing system can only generate simple data curves and integrity level reports in the surface control room, and does not achieve real-time data synchronization between the surface and the downhole (the delay often exceeds 5 seconds). It also lacks visualization (such as 3D rock mass model, fracture spatial distribution) and intelligent early warning functions. Downhole operators need to manually interpret the data reports, which makes it difficult to quickly judge the risk level and adjust the tunneling parameters, resulting in a disconnect between monitoring and decision-making, and failing to effectively prevent sudden geological risks.

[0006] In summary, current rock mass integrity prediction technologies used in TBM tunneling in coal mines suffer from low efficiency, poor accuracy, weak adaptability, and delayed feedback, making it difficult to meet the practical needs of intelligent construction and safety control. Therefore, developing a rock mass integrity prediction method that integrates multi-source data, adapts to new algorithms, and provides real-time visualization of results has become a pressing technical challenge in this field. Summary of the Invention

[0007] The purpose of this invention is to provide a rock integrity prediction method based on drilling while drilling and the CUSUM algorithm. The method collects five types of data in real time—thrust, torque, drilling speed, rotational speed, and vibration—through a drilling while drilling system, and transmits them to the surface control room via fiber optic transmission and a 485 Ethernet port. The system uses a modified CUSUM algorithm to detect fracture locations and combines RQD and Kv indices to evaluate the surrounding rock quality, achieving advanced prediction and intelligent TBM decision-making. This solves the problems of low efficiency, poor accuracy, weak adaptability, and delayed feedback in existing coal mine TBM roadway excavation.

[0008] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: This invention relates to a rock integrity prediction method based on the drilling while drilling and CUSUM algorithm, comprising the following steps: Step S1: Install a drilling data acquisition device on the advanced drilling rig. The data acquired by the drilling data acquisition device includes drilling parameters, elastic wave signals, and real-time video image data of the borehole inner wall; the drilling parameters include thrust, torque, drilling speed, rotational speed, and vibration. Step S2: Process the data collected in step S1; Step S3: Align the processed drilling parameters, acoustic emission events, and fracture images on the time-depth axis; Step S4: The CUSUM algorithm adds a time window detection function, sets an initial drift threshold, monitors whether an offset occurs, and when the cumulative value exceeds the threshold and continues for a preset time, it is determined that a crack exists and the initial offset point is marked as the crack location. Step S5: Calculate the fracture density Jv and rock mass quality index based on the detected fractures, and generate a rock mass quality assessment report; Step S6: Feed the prediction results back to the downhole operator's cab control console in real time, and display the real-time data curve and fracture location markers in the surface control room.

[0009] As a preferred technical solution, in step S1, the data acquisition device while drilling includes a thrust sensor, a torque sensor, a drilling speed sensor, a rotational speed sensor, a vibration sensor, an acoustic emission sensor, and a miniature borehole camera. The thrust sensor is installed in the advanced drilling rig propulsion system to measure the magnitude of the axial thrust during drilling. The torque sensor is installed on the drill rod drive system to monitor the torque change during drilling rig rotation. The drilling speed sensor calculates the drilling speed by measuring the rate of change of axial displacement of the drill rod. The rotational speed sensor is installed on the drilling rig power head or drill rod drive part to measure the rotational speed of the drill rod. The vibration sensor is installed on the drill rod or drilling rig frame to detect vibration signals during drilling. The acoustic emission sensor collects the instantaneous elastic wave signals released by the rock during deformation or fracturing under stress. The miniature borehole camera collects real-time video or image data of the borehole inner wall.

[0010] As a preferred technical solution, in step S2, when processing the drilling parameters collected in step S1, moving average or wavelet threshold denoising is used to unify data from different sampling frequencies onto the same time axis. Identify and remove outlier data using rules or median absolute deviation; extract features from the processed data; When processing the elastic wave signal collected in step S1, wavelet transform or bandpass filtering is used to remove background noise, a threshold is set to identify acoustic emission events, feature extraction is performed on the processed elastic wave signal, a normal drilling acoustic emission baseline model is established, and it is determined whether the rock has developed fractures; when it is determined that the rock has developed fractures, a spatiotemporal correlation analysis is performed with the abnormal detection results of drilling parameters. When processing the video image data acquired in step S1, the video images are first subjected to distortion correction, image enhancement and noise reduction in sequence. The cracks are then identified by a convolutional neural network to determine the location, width and direction of the cracks.

[0011] As a preferred technical solution, the specific process for determining whether rock fractures have developed using the drilling acoustic emission baseline model is as follows: Step S21: Before the TBM tunneling, pilot boreholes are drilled in the rock mass (such as sandstone, mudstone, coal seam) with different lithologies ahead of the tunnel face using an advanced drilling rig. Acoustic emission data (event rate, energy release, main frequency) of each lithology section are collected simultaneously. The collection time for each type of lithology is not less than 30 minutes to ensure that the sample covers the complete working conditions of normal drilling for that lithology. Step S22: For the collected acoustic emission data of normal lithology, divide it into layers according to "drilling stages" (opening section, stable drilling section, and retraction section), and calculate the average event rate (number of acoustic emission events per unit time) for each stage. Standard deviation The average energy release (energy value of a single event) Standard deviation The mean of the dominant frequency (the main frequency component of the acoustic emission signal) Standard deviation This forms a basic statistical feature database based on lithology and stage. Step S23: During normal TBM tunneling, the baseline is updated using a "sliding time window + lithology matching" mechanism; a 5-minute sliding window is set to identify the current drilling lithology in real time (by matching pre-stored lithology database with drilling thrust and torque characteristics), and acoustic emission data without anomaly markers within the window are substituted into the basic statistical feature library, and updated using the exponential weighted average method. This allows the baseline to adapt to subtle changes in geological conditions, avoiding misjudgment of fixed baselines in lithological transition zones. Step S24: Calculate the event rate for the real-time acquired acoustic emission data at 1-second time intervals. Average energy , main frequency Each of these is compared with the corresponding features of the current dynamic baseline: like ( (This is an adjustable coefficient, initially set to 2.5), and is marked as an event rate anomaly. like ( Initially set to 3.0), marked as an energy release anomaly; like ( Initially set to 2.0), marked as a main frequency anomaly; where Optimization can be achieved by backtracking based on historical anomaly data, such as appropriately reducing the k value in areas with high incidence of fractures to improve sensitivity; Step S25: Set the "Abnormal Feature Combination Rule" to determine a potential fracture signal only if any of the following conditions are met: An abnormal event rate and an abnormal energy release occur simultaneously, and the duration is ≥3s; An abnormal event rate and an abnormal main frequency occur simultaneously, and the duration is ≥5s; All three are abnormal at the same time, and the duration is ≥2s; this rule avoids misjudgment caused by a single feature being affected by momentary interference (such as slight jamming of the drill pipe), and improves the specificity of anomaly identification; Step S26: Establish a "drilling depth-time" mapping relationship, using the drill pipe advance speed V (m / min) as the benchmark, and determine the occurrence time of acoustic emission anomalies. Convert to corresponding drilling depth Simultaneously, outliers detected by the CUSUM algorithm in the drilling parameters are also converted into drilling depth. This ensures that the two are aligned in the "drilling depth" dimension; Step S27: For the same drilling depth range The acoustic emission anomalies and drilling parameter anomalies within the drilling system are analyzed, and the correlation coefficient between their characteristics is calculated. Based on the correlation results, the anomaly confidence level is calculated. .

[0012] As a preferred technical solution, the specific process for processing the video image data is as follows: Step T21: Construct a simulated borehole with a diameter of 93mm (consistent with the borehole diameter of the advanced drilling rig) in the laboratory, with a target (containing a grid pattern of known size) inside, collect target images from different angles, and establish a fisheye distortion model; Step T22: Extract lithological features from the image, distinguish lithology by the mean of grayscale histogram, and perform adaptive enhancement according to different lithologies; Step T23: Extract the standard deviation of the preprocessed image for lithology-adaptive Canny edge detection, perform morphological dilation on the initial edge image, connect the fracture edges, calculate the angle between the line segment and the borehole axis, and convert the angle difference... The line segments are classified into the same crack; Step T24: Record the drilling depth and image acquisition angle of the camera device, and establish the mapping relationship between image pixels and actual drilling positions; Step T25: Perform stereo matching on the crack edge and calculate the three-dimensional coordinates of the crack edge based on the principle of triangulation; Step T26: Calculate the orientation angle of the fracture line segment in the image, and correct the fracture direction by combining it with the actual tilt angle of the borehole axis through spatial geometric transformation.

[0013] As a preferred technical solution, in step T21, the fisheye distortion model adopts a polynomial distortion model: ;in The radius after distortion. For the ideal radius, , (where is the distortion coefficient); for borehole images acquired downhole, the ideal coordinates of each pixel are calculated based on a pre-established distortion model, and distortion correction is achieved through bilinear interpolation.

[0014] As a preferred technical solution, the calculation process for crack width and orientation in step T26 is as follows: When calculating the fracture width, the binocular camera device inside the borehole simultaneously acquires left and right images of the fracture area, performs adaptive Gamma correction and Retinex enhancement algorithm, calculates the three-dimensional coordinates of feature points in the left and right images based on the known intrinsic parameters of the binocular camera, and performs preliminary calculation of the fracture width; then, the vibration data corresponding to the fracture depth at the time is extracted, and wavelet packet decomposition is used to extract the energy in the 500Hz frequency band, establishes the relationship model between energy and fracture width, and obtains the final fracture width; When calculating the fracture orientation, an explosion-proof attitude sensor is installed at the end of the drill pipe of the TBM-mounted advanced drilling rig to collect the inclination angle and azimuth angle of the borehole in real time. The acquisition frequency is synchronized with the image inside the borehole to obtain the attitude parameters corresponding to the fracture depth. The fracture angle in the image plane is extracted, and a transformation model between the borehole local coordinate system and the downhole geodetic coordinate system is established. The fracture angle in the image plane is converted into the orientation angle in the geodetic coordinate system through Euler angle transformation, and finally the fracture orientation is determined.

[0015] As a preferred technical solution, the specific process for aligning drilling parameters, acoustic emission events, and fracture images on the time-depth axis in step S3 is as follows: Step S31: Acquire drilling parameters, acoustic emission events, and fracture images. Using the borehole start time as the time origin and the initial drill bit position as the depth origin, perform integral calculation of the depth using the drilling rate sensor. In the formula, for Drilling speed at any moment for Drilling depth at any given time; Step S32: Drilling parameters are directly mapped to the absolute time axis; acoustic emission events are time-stamped, and the trigger time of each event is recorded; fracture images are synchronously recorded with the current drilling depth based on the absolute timestamp of the capture time; Step S33: Use the modified formula Dynamic error compensation is performed, where, The elastic coefficient, for The thrust value of the drill pipe at any given time; Step S34: For the nth frame image, record its corresponding depth. , This refers to the fixed distance between the camera device and the drill bit; Step S35: Record the start time of the anomaly detected by the USUM algorithm. Corresponding depth ; Filter out time windows Acoustic emission events within the range of depth Depth retrieval in image sequences The corresponding image frame is used to identify the crack features in that frame and the two adjacent frames.

[0016] As a preferred technical solution, in step S5, the crack density is calculated based on the detected crack conditions. And the rock mass quality indicators, the specific formulas are as follows: ;based on The values ​​are substituted into the formula to calculate the rock quality indicators respectively. and fracture volume factor The specific formula is as follows: , .

[0017] As a preferred technical solution, in step S6, the prediction results are fed back to the control console in the downhole operator's cab in real time. The main screen displays a three-dimensional rock mass integrity model. Based on the Kriging interpolation algorithm, different colors are used to render the rock mass quality within 50m in front of the tunnel face, with red representing the extremely fractured area and green representing the intact area. The secondary screen 1 displays real-time data curves, and the secondary screen 2 displays real-time images of the borehole.

[0018] The present invention has the following beneficial effects: This invention collects five types of data in real time—thrust, torque, drilling speed, rotational speed, and vibration—through a drilling system, and transmits them via fiber optic cable and 485 Ethernet port to the surface control room. The system uses a modified CUSUM algorithm to detect fracture locations and combines RQD and Kv indices to evaluate the surrounding rock quality, achieving advanced prediction and TBM intelligent decision-making.

[0019] This invention breaks through the traditional depth calculation method of "single drilling speed integral" and establishes a three-dimensional spatiotemporal anchoring model of "drilling parameter timestamp - in-hole image frame - drill pipe displacement". Combined with dynamic compensation for drilling speed fluctuations, it achieves sub-millimeter depth positioning.

[0020] This invention not only uses five data points—thrust, torque, drilling speed, rotational speed, and vibration—but also introduces acoustic emission (AE) signals and borehole images (acquired through a miniature camera probe). These heterogeneous data are then fused. Acoustic emission can detect microcrack initiation earlier and is more sensitive than vibration signals. Borehole images can intuitively verify the algorithm's prediction results, reducing misjudgments. A feature-level fusion model is established, and an attention mechanism is used to allow the model to automatically weigh the contributions of different sensors.

[0021] This invention automatically adjusts the drift threshold and window length of CUSUM based on different lithologies (determined through historical borehole data and lithology databases), incorporates a sliding learning rate, reduces sensitivity in stable strata, and increases sensitivity in complex strata, avoiding false alarms triggered by random noise in homogeneous strata, and enabling faster anomaly detection in fractured areas.

[0022] This invention first identifies lithological types (such as sandstone, mudstone, and coal seams) through drilling data, and then adopts customized integrity assessment formulas for different lithologies. It establishes a mapping relationship between RQD and Kv and rock mass strength parameters to form a multi-dimensional rock mass quality index (RMI), which reflects rock mass stability more comprehensively than simply using RQD / Kv, and can directly guide the dynamic optimization of TBM tunneling parameters (thrust, rotation speed).

[0023] This invention uses spatial interpolation algorithms to generate a three-dimensional rock mass integrity model in front of the tunnel face from the prediction results of multiple boreholes; it displays a real-time updated three-dimensional geological cross-section in the TBM operator's cab, which facilitates the operator's intuitive understanding of the geological structure ahead; and it provides a more intuitive basis for the selection of support schemes.

[0024] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart of a rock integrity prediction method based on drilling and CUSUM algorithms according to the present invention. Figure 2 A schematic diagram of the rock integrity prediction device. Figure 3 This diagram illustrates the processes of data processing while drilling, fracture identification, and rock integrity prediction. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0029] To make the purpose, technical solution, and advantages of this application clearer, the following description is provided in conjunction with the appendix. Figure 1-3The present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the application.

[0030] Example 1

[0031] Please see Figure 1 As shown, this invention is a rock integrity prediction method based on the drilling while drilling and CUSUM algorithm, comprising the following steps: Step S1: Install a drilling data acquisition device on the advanced drilling rig. The data collected by the drilling data acquisition device includes drilling parameters, elastic wave signals, and real-time video image data of the borehole inner wall; the drilling parameters include thrust, torque, drilling speed, rotational speed, and vibration. Step S2: Process the data collected in step S1; Step S3: Align the processed drilling parameters, acoustic emission events, and fracture images on the time-depth axis; Step S4: The CUSUM algorithm adds a time window detection function, sets an initial drift threshold, monitors whether an offset occurs, and when the cumulative value exceeds the threshold and continues for a preset time, it is determined that a crack exists and the initial offset point is marked as the crack location. Step S5: Calculate the fracture density Jv and rock mass quality index based on the detected fractures, and generate a rock mass quality assessment report; Step S6: Feed the prediction results back to the downhole operator's cab control console in real time, and display the real-time data curve and fracture location markers in the surface control room.

[0032] In step S2, the drilling parameters acquired in step S1 are denoised using moving average or wavelet thresholding to unify data from different sampling frequencies onto the same time axis. Identify and remove outlier data using rules or median absolute deviation; extract features from the processed data; Specifically, the moving average (window size 5) Denoising is performed using 20 sampling points or wavelet thresholding (db4 wavelet basis, decomposed into 5 layers) to eliminate transient pulse interference (such as sensor jitter, motor start-up impact); signal trends and true changes are preserved, reducing false detections caused by noise; during time synchronization, the thrust, torque, drilling speed, and rotational speed are resampled linearly using the vibration signal with the highest sampling frequency as the reference; the sampling interval is unified to 10ms to ensure that the five signals are aligned at the same time point, facilitating subsequent multi-parameter joint analysis; when removing outliers, the mean value within the sliding window of each parameter is calculated. and standard deviation Exceeding Values ​​outside this range are considered outliers; Feature extraction of the processed data includes: time-domain features, frequency-domain features, and statistical features. For time-domain feature extraction, for each parameter, the following are calculated within a 50-200ms sliding window: mean (reflecting the overall level), variance (reflecting the degree of fluctuation), peak value (maximum instantaneous value), kurtosis (reflecting the signal spike), waveform factor (RMS / mean, reflecting the waveform shape), and trend slope (the slope of the linear fit within the window, reflecting the trend of change). For frequency-domain feature extraction, an FFT (Fast Fourier Transform) is performed on the window data; the dominant frequency (the frequency with the highest energy) and spectral energy distribution (the proportion of energy in each frequency band) are extracted. For vibration signals, the focus is on the 200ms level. Variations in the 1000Hz frequency band often reflect the degree of rock mass fracturing.

[0033] When processing the elastic wave signal collected in step S1, wavelet transform or bandpass filtering is used to remove background noise, a threshold is set to identify acoustic emission events, feature extraction is performed on the processed elastic wave signal, a normal drilling acoustic emission baseline model is established, and it is determined whether the rock has developed fractures; when it is determined that the rock has developed fractures, a spatiotemporal correlation analysis is performed with the abnormal detection results of drilling parameters. When processing the video image data acquired in step S1, the video images are first subjected to distortion correction, image enhancement and noise reduction in sequence. The cracks are then identified by a convolutional neural network to determine the location, width and direction of the cracks.

[0034] Please see Figure 3 As shown, the specific process for determining whether rock fractures have developed using the drilling acoustic emission baseline model is as follows: Step S21: Before the TBM tunneling, pilot boreholes are drilled in the rock mass (such as sandstone, mudstone, coal seam) with different lithologies ahead of the tunnel face using an advanced drilling rig. Acoustic emission data (event rate, energy release, main frequency) of each lithology section are collected simultaneously. The collection time for each type of lithology is not less than 30 minutes to ensure that the sample covers the complete working conditions of normal drilling for that lithology. Step S22: For the collected acoustic emission data of normal lithology, divide it into layers according to "drilling stages" (opening section, stable drilling section, and retraction section), and calculate the average event rate (number of acoustic emission events per unit time) for each stage. Standard deviation The average energy release (energy value of a single event) Standard deviation The mean of the dominant frequency (the main frequency component of the acoustic emission signal) Standard deviation This forms a basic statistical feature database based on lithology and stage. Step S23: During normal TBM tunneling, the baseline is updated using a "sliding time window + lithology matching" mechanism; a 5-minute sliding window is set to identify the current drilling lithology in real time (by matching pre-stored lithology database with drilling thrust and torque characteristics), and acoustic emission data without anomaly markers within the window are substituted into the basic statistical feature library, and updated using the exponential weighted average method. This allows the baseline to adapt to subtle changes in geological conditions, avoiding misjudgment of fixed baselines in lithological transition zones. Step S24: Calculate the event rate for the real-time acquired acoustic emission data at 1-second time intervals. Average energy , main frequency Each of these is compared with the corresponding features of the current dynamic baseline: like ( (This is an adjustable coefficient, initially set to 2.5), and is marked as an event rate anomaly. like ( Initially set to 3.0), marked as an energy release anomaly; like ( Initially set to 2.0), marked as a main frequency anomaly; where Optimization can be achieved by backtracking based on historical anomaly data, such as appropriately reducing the k value in areas with high incidence of fractures to improve sensitivity; Step S25: Set the "Abnormal Feature Combination Rule" to determine a potential fracture signal only if any of the following conditions are met: An abnormal event rate and an abnormal energy release occur simultaneously, and the duration is ≥3s; An abnormal event rate and an abnormal main frequency occur simultaneously, and the duration is ≥5s; All three are abnormal at the same time, and the duration is ≥2s; this rule avoids misjudgment caused by a single feature being affected by momentary interference (such as slight jamming of the drill pipe), and improves the specificity of anomaly identification; For signals identified as potential fractures, calculate the gradient values ​​of the anomalous features (such as the increase in the event rate within 5 seconds). The fluctuation range of energy release ,according to The size classifies the abnormality level into "minor" ( , "Moderate ( , )","serious( , This provides a quantitative basis for subsequent integrity assessment; Step S26: Establish a "drilling depth-time" mapping relationship, using the drill pipe advance speed V (m / min) as the benchmark, and determine the occurrence time of acoustic emission anomalies. Convert to corresponding drilling depth Simultaneously, outliers detected by the CUSUM algorithm in the drilling parameters are also converted into drilling depth. Ensure that the two are aligned in the "drilling depth" dimension (error controlled within ±0.1m). Step S27: For the same drilling depth range The correlation coefficient between acoustic emission anomalies and drilling parameter anomalies is calculated. If the correlation coefficient between acoustic emission event rate anomalies and drilling torque anomalies is high... Furthermore, the correlation coefficient between abnormal energy release and abnormal drilling vibration is... The result was determined to be a "strong correlation anomaly," confirming the presence of fractures in the drilling depth section. If an anomaly is detected only from a single data source (acoustic emission or drilling parameters), or the correlation coefficient... The image is marked as "abnormal and needs to be verified", which triggers the subsequent in-hole image verification process. Calculate the anomaly confidence level based on the association results. The formula is: ,in Assign a confidence level to the acoustic emission anomaly (based on the anomaly level: slight 0.4, moderate 0.7, severe 0.9). The confidence level of the drilling parameter anomalies is determined by the percentage of the cumulative offset detected by the CUSUM algorithm relative to the threshold, ranging from 0 to 1. The average correlation coefficient for multiple features. The weighting coefficients (calibrated to 0.4, 0.4, and 0.2 through field tests) are used when... If the condition is not found, it is ultimately determined to be a fracture development; otherwise, further verification is required by combining images from inside the borehole.

[0035] In practice, during the tunneling process, the average event rate of identifying the current lithology as medium-grained sandstone within a sliding window (5 minutes) with no abnormal data annotations is [value missing]. Average energy release 78 , main frequency The baseline is updated using an exponentially weighted average (weight 0.7): , , .

[0036] When the drilling depth reached 125.3m, the real-time data within a 1-second interval was: event rate. ( Energy release ( ), main frequency ( If the event rate is abnormal and the energy release is abnormal, and the condition persists for 4 seconds, it is considered a potential fracture signal, and the abnormality level is "severe". , ), .

[0037] Acoustic emission anomaly drilling depth The CUSUM algorithm for drilling parameters in An abnormal torque was detected (cumulative offset as a percentage of the threshold is 0.8). ); Calculate the correlation coefficient: event rate and torque anomaly Energy release and vibration anomalies , Confidence level Final judgment There are cracks in the deep section of the drilled section.

[0038] Verification results: Subsequent borehole images clearly showed a longitudinal crack with a width of about 2 mm at 125.25 m, which was consistent with the acoustic emission-while-drilling parameter correlation results, verifying the effectiveness of the method.

[0039] The specific process for processing video image data is as follows: Step T21: Construct a simulated borehole with a diameter of 93mm (consistent with the borehole diameter of the advanced drilling rig) in the laboratory, with an embedded target (containing a grid pattern of known dimensions). Acquire target images from different angles to establish a fisheye distortion model (using a polynomial distortion model). ,in The radius after distortion. For the ideal radius, , (This refers to the distortion coefficient). In practice, for borehole images acquired downhole, the ideal coordinates of each pixel are calculated based on a pre-established distortion model. Distortion correction is achieved through bilinear interpolation to ensure accuracy in measuring fracture size in the image. ; Step T22: Extract lithological features from the image, using the mean of the grayscale histogram (... Differentiate lithology and perform adaptive enhancement based on different lithologies, typically sandstone. Usually mudstone for coal seam for ; Specifically, if it is sandstone ( ): Limit contrast adaptive histogram equalization (CLAHE) is used, with the clip limit set to 2.0, to enhance the grayscale difference between the fracture and the rock matrix; if it is mudstone / coal seam ( First, perform gamma correction ( Increase the overall brightness and then perform CLAHE (clip limit=1.5) to prevent cracks in low grayscale images from being masked by noise; Step T23: Extract the standard deviation of the preprocessed image for lithology-adaptive Canny edge detection, perform morphological dilation on the initial edge image, connect the fracture edges, calculate the angle between the line segment and the borehole axis, and convert the angle difference... The line segments are classified into the same crack; In practice, the standard deviation of grayscale values ​​in the preprocessed image is extracted. ):sandstone Usually mudstone for coal seam for High threshold low threshold (in sandstone) , mudstone , ); The Canny operator is used for edge detection to obtain an initial fracture edge image, avoiding the failure to detect fractures in low-contrast lithology (such as coal seams) or the false detection of impurity edges in high-contrast lithology (such as sandstone) by using a fixed threshold. Step T24: Record the drilling depth of the camera device (synchronously acquired by the drilling speed sensor, denoted as D) and the image acquisition angle (recorded by the built-in gyroscope of the camera device, denoted as θ, 0° is directly above the drilling hole, increasing clockwise), and establish the mapping relationship between image pixels and actual drilling position; Specifically, establish the mapping relationship between image pixels and actual drilling locations: the number of pixels in the horizontal (circumferential) direction of the image is... Corresponding to the circumference of the drill hole Then each pixel corresponds to the actual circumference length. The number of pixels in the vertical direction (axial direction) of the image is The corresponding axial length of a single acquisition by the camera device (Based on the lens field of view setting), each pixel corresponds to the actual axial length. ; If the crack starts at pixel 1 in the horizontal direction in the image The terminating pixel is The starting pixel in the vertical direction is The terminating pixel is The actual axial position of the crack is (Start) to (Termination), circumferential position is (Start) to (termination); Step T25: Perform stereo matching on the crack edge (use SIFT algorithm to extract feature points and calculate disparity), calculate the three-dimensional coordinates of the crack edge based on the principle of triangulation; calculate the shortest distance between the two sides of the crack edge in the direction perpendicular to the crack direction, which is the actual width of the crack; Step T26: Calculate the orientation angle of the fracture line segment in the image, and correct the fracture direction by combining it with the actual tilt angle of the borehole axis through spatial geometric transformation.

[0040] Specifically, the fracture orientation is corrected through spatial geometric transformation by combining the actual inclination angle of the borehole axis (obtained by the TBM attitude sensor and denoted as β, i.e., the angle between the borehole and the horizontal direction): actual orientation angle. (like (Then subtract 90°). Cracks are classified according to their orientation angle: It is a "gently tilting fracture". It is a "medium-dipping fracture". The term "steeply inclined fracture" provides more detailed parameters for subsequent rock mass stability assessment.

[0041] In practice, the drilling parameters were: thrust 65 kN, torque 12 kN·m, drilling speed 0.4 m / min, rotational speed 25 rpm, vibration acceleration 3.5 g, and sampling rate 100 Hz; the acoustic emission data were: sampling rate 1 MHz, event rate 8 times / s (5-10 times / s in the sandstone baseline range), and average energy. (baseline interval) ), main frequency 180kHz (baseline range) ); The images inside the hole were acquired using a binocular miniature camera device, with an image resolution of 1280×720 and an acquisition depth of [missing information]. ,angle .

[0042] When drilling reached a depth of 150.2 m, the acoustic emission event rate surged to 18 events / s (exceeding the upper limit of 15 events / s for the sandstone baseline) and the average energy increased to 28 μV·s (exceeding the upper limit of 25 μV·s for the baseline), with a characteristic change gradient of 40%, which was marked as a "fracture development signal"; The CUSUM algorithm detected a continuous torque shift (from 12 kN·m to 14 kN·m) in the drilling parameters within the depth range, and the fusion judgment was "confirmed fracture".

[0043] The specific implementation process for calculating the crack depth is as follows: Multi-source data timestamp synchronization: A high-precision time synchronization module (accuracy ≤ 1ms) is used to unify the timestamps for thrust, torque, drilling speed, rotational speed, vibration, and borehole image data, denoted as […]. The drilling speed sensor collects instantaneous drilling speed in real time. (Unit: mm / s), the intra-hole camera device per Capture one frame of image, denoted as ( (Image frame number); Calculate the initial depth by integrating the drilling rate. ( (This refers to the borehole start time); using feature points of drilling parameters for anchoring, the fracture offset start point of drilling parameters (such as torque) is detected by the CUSUM algorithm (the core algorithm of the original file). Corresponding to the initial depth ( );exist Extract images inside the well within ±0.2s before and after. Identify natural markers (such as fixed textures and micro-protrusions) on the inner wall of boreholes in images and calculate the axial displacement of these markers in the image. (pixels); based on image calibration parameters (laboratory pre-calibration: number of image pixels per millimeter along the borehole axis). ), convert to actual axial displacement The initial depth is corrected to: ("+" and "-" are determined based on the direction of displacement of the marker); Perform dynamic compensation for drilling rate fluctuations and calculate Drilling rate fluctuation coefficient within 1 second before and after the time: ;like (Significant drilling rate fluctuations) A thrust-drilling rate correlation compensation model is introduced: based on the thrust data F( collected from the original file) The polynomial fitting formula trained using a historical database. ( The lithological correlation coefficient is for sandstone. mudstone ), calculate the compensated drilling speed The final fracture depth is The accuracy can reach ±0.3mm.

[0044] When calculating the fracture width, a binocular camera device inside the borehole (baseline distance B=15mm, resolution 1280×720) simultaneously acquires left and right images of the fracture area. For the low-light environment of underground coal mines, adaptive Gamma correction and Retinex enhancement algorithms are implemented; specifically, the average grayscale value of the image is calculated. ,like (Low light), set ;like ,set up ;like ,set up The Retinex algorithm is used to separate the "illuminance component" and "reflectance component" of the image, suppressing grayscale deviations caused by uneven illumination; the pre-processed image... SIFT feature points are extracted from the crack edges, and mismatched points are removed using the RANSAC algorithm, retaining the matching pairs. Based on known intrinsic parameters of the binocular camera (laboratory calibration: focal length) pixel size Calculate the feature points of the left and right images. The three-dimensional coordinates are used to initially calculate the crack width; specifically, the left camera coordinate system is... ( (The coordinates of the principal point of the camera are shown); the coordinate system of the right camera is shown. Constrained by binocular baseline The depth of the feature points is obtained. This allows us to obtain the three-dimensional coordinates of the edge points on both sides of the crack. The rough estimate of the width is... .

[0045] Extracting crack depth The vibration data at the corresponding time (one of the five types of drilling data in the original file) was used to extract the energy in the 100-500Hz frequency band using wavelet packet decomposition. (This frequency band is the characteristic vibration frequency band generated by fracture collision); then, the vibration data corresponding to the fracture depth at that time is extracted, and wavelet packet decomposition is used to extract the energy of the 500Hz frequency band. A relationship model between energy and fracture width is established to obtain the final fracture width. The correlation model with crack width was obtained by training on historical data. ( For correction factor, Unit: 10^-3 V²·s, k=0.02); final crack width is Measurement error ≤ ±0.05mm (better than ±0.2mm for monocular measurement); When calculating the fracture orientation, an explosion-proof attitude sensor is installed at the end of the drill rod of the TBM-mounted advanced drilling rig to collect the inclination angle and azimuth angle of the borehole in real time. The acquisition frequency is synchronized with the borehole image to obtain the attitude parameters corresponding to the fracture depth. The fracture angle in the image plane is extracted, and a transformation model between the borehole local coordinate system and the downhole geodetic coordinate system is established. The fracture angle in the image plane is converted into the orientation angle in the geodetic coordinate system through Euler angle transformation, and finally the fracture orientation is determined. In practice, an explosion-proof attitude sensor is installed at the end of the drill rod of the TBM-mounted advanced drilling rig to collect the borehole tilt angle in real time. The angle between the borehole axis and the horizontal direction ( ); Azimuth The angle between the projection of the borehole axis onto the horizontal plane and the due north direction ( The acquisition frequency is synchronized with the image inside the hole. Record the fracture depth. The corresponding attitude parameters are ; Image of the inside of the hole (Img) An improved Canny-Hough transform is used to extract fracture edges: Canny edge detection: based on lithology-adaptive thresholds (sandstone: high threshold Th=50, low threshold Tl=20; mudstone: Th=35, Tl=15), avoiding missed / false detections with fixed thresholds; Hough transform: setting a minimum line segment length. (Corresponding to image pixel 25), maximum gap =1mm (corresponding to 5 pixels), detect the line segment direction angle at the crack edge. (In the image plane, the angle with the vertical direction is positive when clockwise).

[0046] Establish a transformation model between the borehole local coordinate system and the downhole geodetic coordinate system; specifically, the borehole local coordinate system O'-X'Y'Z': O' is the center point of the fracture, the X' axis points into the borehole along the borehole axis, the Y' axis is vertically upward within the borehole cross-section, and the Z' axis is horizontally to the right within the cross-section; the downhole geodetic coordinate system O-XYZ: the X axis is due north, the Y axis is due east, and the Z axis is vertically upward; Through Euler angle transformation (rotating sequentially around the Z-axis) Rotation around the Y-axis ), the crack angle in the image plane Strike angle converted to geodetic coordinates : like (Horizontal drilling), then (like (reduced by 360°) like ≠0° (inclined drilling), then (Correcting angular deviations caused by tilting through spatial projection); Final fracture orientation according to The division (0°-60° is northeast, 60°-120° is east, 120°-180° is southeast, and so on) can be accurate to ±2°.

[0047] In step S3, the specific process for aligning drilling parameters, acoustic emission events, and fracture images on the time-depth axis is as follows: Step S31: Acquire drilling parameters, acoustic emission events, and fracture images. Using the borehole start time as the time origin and the initial drill bit position as the depth origin, perform integral calculation of the depth using the drilling rate sensor. In the formula, for Drilling speed at any moment for The drilling depth at any given moment; a synchronization signal is triggered when drilling begins, so that the time axis of the acoustic emission sensor, the in-hole camera device, and the drilling parameter acquisition system are unified to an absolute timestamp (accurate to the millisecond level). Step S32: Drilling parameters are directly mapped to the absolute time axis; acoustic emission events are time-stamped, and the trigger time of each event is recorded; fracture images are synchronously recorded with the current drilling depth based on the absolute timestamp of the capture time; Step S33: Due to the depth measurement deviation caused by the elastic deformation of the drill pipe, a correction formula is used. Dynamic error compensation is performed, where, The elastic coefficient, for The thrust value of the drill pipe at any given time. To correct the actual borehole depth, The original measured depth is calculated based on the drilling rate integral. Step S34: For the nth frame image, record its corresponding depth. , This refers to the fixed distance between the camera device and the drill bit; Step S35: Record the start time of the anomaly detected by the USUM algorithm. Corresponding depth ; Filter out time windows Acoustic emission events within the range of depth Depth retrieval in image sequences The corresponding image frame is used to identify the crack features in that frame and the two adjacent frames.

[0048] If anomalies are detected in all three types of data within a depth range of ±50mm, it is classified as a "strongly consistent crack" with a weight of 0.8; if anomalies are detected in two types of data, it is classified as a "moderately consistent crack" with a weight of 0.5; if anomalies are detected in only one type of data, it is classified as a "weakly consistent crack" with a weight of 0.2 (manual verification required).

[0049] In practice, at the 10th second of drilling, the drilling speed v = 100 mm / s, and the calculated depth h = 100 × 10 = 1000 mm; at the 15th second, the thrust F = 50 kN, the elastic coefficient k = 0.02 mm / kN, and the corrected depth is: =100×15-0.02×50=1500-1=1499mm; the drilling parameters triggered a CUSUM anomaly at t=20s (thrust suddenly increased from 30kN to 45kN), corresponding to a depth h=2000mm; three high-energy acoustic emission events (amplitude >80dB) were detected between t=19.9-20.1s, corresponding to depths of 1990-2010mm, with a depth deviation of <10mm from the anomaly point; the image frame at t=20s corresponds to a depth of h=2000+500=2500mm (camera device 500mm from drill bit), and a 0.5mm wide fracture was identified in this frame. Based on the drilling rate, the actual location of the fracture was calculated to be 2500-500=2000mm, which perfectly matches the anomaly point; anomalies were detected at a depth of 2000mm in all three types of data, which were determined to be "strongly consistent fractures" and included in the rock mass integrity assessment results.

[0050] In step S5, the crack density is calculated based on the detected crack conditions. And the rock mass quality indicators, the specific formulas are as follows: ;based on The values ​​are substituted into the formula to calculate the rock quality indicators respectively. and fracture volume factor The specific formula is as follows: , .

[0051] Specifically, The calculation uses "per meter of borehole section" as the statistical unit. Multiple data sources are used to confirm the specific number of fractures, prioritizing 1-meter continuous sections (e.g., 10m-11m, 11m-12m). If the borehole length is less than 1 meter (e.g., the final section 19.6m-20m), it is calculated based on the actual length (e.g., a 0.4m section of 19.6m-20m with two detected fractures, then Jv = 2 / 0.4 = 5 fractures / m). When different data (e.g., drilling vibration, borehole images) detect the same fracture at the same depth, it is counted only once. If multiple parallel fractures exist at the same depth (spacing < 0.05m), they are counted as one fracture (to avoid duplicate counting). Example: A 1m borehole section (15m-16m) is confirmed by drilling parameters, acoustic emission, and borehole images to have 3 fractures (located at 15.2m, 15.5m, and 15.8m respectively). .

[0052] For every additional Jv / m, A decrease of approximately 0.0236 ensures and They exhibit a linear negative correlation, and when Jv=0 (no cracks), (The actual upper limit is set to 1.0, representing a complete rock mass); when hour, (Representing extremely fractured rock mass), covering common geological scenarios in coal mine TBM tunneling.

[0053] Example: If ,but The actual value is 1.0 (for intact rock mass); if ,but (Corresponding to the "intact" rock mass integrity level).

[0054] like and The levels of judgment are consistent (e.g.) Both correspond to "good / relatively complete"), directly output the level; if there is a discrepancy between the two judgments (e.g. Corresponding to "good", Corresponding to "General"), the actual density of cracks in the image inside the hole is used as the standard for correction (if the image shows 5 cracks per meter, which is biased towards "General", then the final grade is "General / Relatively Broken").

[0055] In step S6, the prediction results are fed back to the control console in the downhole operator's cab in real time. The main screen displays a three-dimensional rock mass integrity model. Based on the Kriging interpolation algorithm, different colors are used to render the rock mass quality within 50m in front of the tunnel face, with red representing the extremely fractured zone and green representing the intact zone. The secondary screen 1 displays real-time data curves, and the secondary screen 2 displays real-time images of the borehole.

[0056] Specifically, the system hardware architecture is as follows: The underground driver's cab control console and the above-ground TBM control room are connected by a dual-link transmission channel consisting of a mine fiber optic communication network and an industrial Ethernet switch. The main link uses single-mode fiber (transmission rate ≥1000Mbps), and the backup link uses an intrinsically safe mine-use wireless base station (supporting 4G / 5G private network, bandwidth ≥100Mbps) to ensure transmission redundancy.

[0057] The underground terminal uses a mining-grade explosion-proof and intrinsically safe touchscreen (size ≥ 10.1 inches, resolution 1920×1200, protection level IP67), with a built-in industrial-grade processor (main frequency ≥ 2.0GHz, memory ≥ 4GB), supporting real-time data caching (local storage capacity ≥ 128GB) to avoid data loss during network interruptions. It adopts a two-level architecture of "edge caching + cloud storage": the underground edge computing node (mining explosion-proof server) caches all data within 72 hours in real time (including raw acquisition data, preprocessing results, and prediction reports), and the surface database (using a MySQL cluster, storage capacity ≥ 10TB) automatically synchronizes edge node data every morning, and stores it in partitions according to "year-month-project section"; it provides a "multi-condition combination query" interface, supporting data filtering by time (accurate to the second), drilling depth (accurate to 0.1m), rock mass integrity level, fracture type, etc.; query results can be exported to Excel (including parameter values ​​and statistical indicators), PDF (including curve charts and evaluation reports), and CSV (raw data) formats, with an export speed ≤ 100MB / minute.

[0058] It supports offline analysis interfaces and can be connected to tools such as MATLAB and Python. The exported data includes a data format specification document (including parameter units, sampling frequency, and coordinate mapping rules).

[0059] Based on historical data and on-site engineering experience, establish tiered early warning thresholds: Level 1 Warning (Emergency): or crack density within 10 seconds An increase of ≥5 lines / meter, or a sudden increase of ≥100 acoustic emission events / minute; Level 2 Warning (Reminder): or crack density within 30 seconds Increase by ≥3 strips / meter; Level 3 Warning (Attention): or the crack density within 1 minute Increase by ≥2 strips / meter.

[0060] The threshold can be dynamically adjusted; administrators can lower the warning level by one level based on current geological conditions (e.g., when the lithology is mudstone). The threshold can be manually modified (up to 20), or automatically optimized by the system based on data from the past hour (using the sliding window mean method to update the threshold).

[0061] The rock quality grading standards are as follows:

[0062] Example 2 See Figure 2As shown, a specific apparatus for implementing a rock integrity prediction method can be used to execute the method content of Embodiment 1 of the present invention, including a wellhead control room, a driver's cab control console, and a drilling system installed on the advanced drilling rig; The drilling system includes thrust sensors, torque sensors, drilling speed sensors, vibration sensors, acoustic emission sensors, and a miniature in-hole camera. Specifically: the thrust sensor collects thrust data and is installed in the advance drilling rig propulsion system to measure the magnitude of axial thrust during drilling; the torque sensor collects torque data and is installed on the drill pipe drive system to monitor torque changes during drill rig rotation; the drilling speed sensor (or displacement sensor) collects drilling speed data and calculates the drilling speed by measuring the rate of change of axial displacement of the drill pipe; the rotational speed sensor (such as a Hall sensor or encoder) collects rotational speed data and is installed in the drill pipe drive system. Mounted on the drill rig's power head or drill pipe drive unit, it measures the rotational speed of the drill pipe; vibration sensors (accelerometers) collect vibration data and are mounted on the drill pipe or drill rig frame to detect vibration signals during drilling, especially through fiber optic sensors to adapt to the downhole environment; acoustic emission sensors collect instantaneous elastic wave signals (acoustic emission signals) released by rocks during stress deformation or fracturing, which can be used to detect the initiation and propagation of microcracks at an early stage; micro-hole camera devices collect real-time video or image data of the borehole inner wall, allowing for direct observation of rock strata structure, fracture distribution, lithological changes, etc. Data collected by the drilling system is uploaded to the operator's cab control console via a junction box. The operator's cab control console then transmits the data to the surface storage and processes and predicts it using the CUSUM algorithm. The surface control room then feeds back the predicted results to the downhole operator's cab control console to assist in the TBM's intelligent decision-making.

[0063] It is worth noting that the various units included in the above system embodiments are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0064] Furthermore, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware, and the corresponding program can be stored in a computer-readable storage medium.

[0065] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A rock integrity prediction method based on drilling while drilling and CUSUM algorithm, characterized in that, Includes the following steps: Step S1: Install a drilling data acquisition device on the advanced drilling rig. The data acquired by the drilling data acquisition device includes drilling parameters, elastic wave signals, and real-time video image data of the borehole inner wall; the drilling parameters include thrust, torque, drilling speed, rotational speed, and vibration. Step S2: Process the data collected in step S1; Step S3: Align the processed drilling parameters, acoustic emission events, and fracture images on the time-depth axis; Step S4: The CUSUM algorithm adds a time window detection function, sets an initial drift threshold, monitors whether an offset occurs, and when the cumulative value exceeds the threshold and continues for a preset time, it is determined that a crack exists and the initial offset point is marked as the crack location. Step S5: Calculate the fracture density Jv and rock mass quality index based on the detected fractures, and generate a rock mass quality assessment report; Step S6: Feed the prediction results back to the downhole operator's cab control console in real time, and display the real-time data curve and fracture location markers in the surface control room.

2. The rock integrity prediction method based on drilling and CUSUM algorithm according to claim 1, characterized in that, In step S1, the data acquisition device while drilling includes a thrust sensor, a torque sensor, a drilling speed sensor, a rotational speed sensor, a vibration sensor, an acoustic emission sensor, and a miniature borehole camera. The thrust sensor is installed in the advance drilling rig propulsion system to measure the magnitude of the axial thrust during drilling. The torque sensor is installed on the drill pipe drive system to monitor the torque change during drilling rig rotation. The drilling speed sensor calculates the drilling speed by measuring the rate of change of axial displacement of the drill pipe. The rotational speed sensor is installed on the drilling rig power head or drill pipe drive part to measure the rotational speed of the drill pipe. The vibration sensor is installed on the drill pipe or drilling rig frame to detect vibration signals during drilling. The acoustic emission sensor collects the instantaneous elastic wave signals released by the rock during deformation or fracturing under stress. The miniature borehole camera collects real-time video or image data of the borehole inner wall.

3. The rock integrity prediction method based on drilling and CUSUM algorithm according to claim 1, characterized in that, In step S2, when processing the drilling parameters acquired in step S1, moving average or wavelet threshold denoising is used to unify data from different sampling frequencies onto the same time axis. Identify and remove outlier data using rules or median absolute deviation; extract features from the processed data; When processing the elastic wave signal collected in step S1, wavelet transform or bandpass filtering is used to remove background noise, a threshold is set to identify acoustic emission events, feature extraction is performed on the processed elastic wave signal, a normal drilling acoustic emission baseline model is established, and it is determined whether the rock has developed cracks. Once it is determined that the rock has developed fractures, a spatiotemporal correlation analysis is performed with the results of abnormal drilling parameter detection. When processing the video image data acquired in step S1, the video images are first subjected to distortion correction, image enhancement and noise reduction in sequence. The cracks are then identified by a convolutional neural network to determine the location, width and direction of the cracks.

4. The rock integrity prediction method based on drilling and CUSUM algorithm according to claim 3, characterized in that, The specific process for using the drilling acoustic emission baseline model to determine whether rock fractures have developed is as follows: Step S21: Before the TBM tunneling, pilot boreholes of different rock types in front of the tunnel face are drilled using an advanced drilling rig, and acoustic emission data of each rock type section are collected simultaneously. Step S22: For the collected normal lithological acoustic emission data, divide the data into layers according to drilling stages and calculate the average event rate for each stage. Standard deviation Mean energy release Standard deviation average clock speed Standard deviation This forms a basic statistical feature database based on lithology and stage. Step S23: During normal TBM tunneling, the baseline is updated using a sliding time window and lithology matching mechanism; A 5-minute sliding window is set to identify the current drilling lithology in real time. Acoustic emission data without anomaly markers within the window are then substituted into the basic statistical feature database, and updated using an exponential weighted average method. This allows the baseline to adaptively adjust to changes in geological conditions. Step S24: Calculate the event rate for the real-time acquired acoustic emission data at 1-second time intervals. Average energy , main frequency Each of these is compared with the corresponding features of the current dynamic baseline: If If so, it is marked as an event rate anomaly; if If so, it is marked as an abnormal energy release; like If it is, then it is marked as a main frequency anomaly; where Optimization can be achieved by backtracking based on historical anomaly data; Step S25: Set abnormal feature combination rules to determine whether it is a potential fracture signal; Step S26: Establish the drilling depth-time mapping relationship, based on the drill pipe advance speed. Based on the time of occurrence of acoustic emission anomalies, Convert to corresponding drilling depth ; Step S27: For the same drilling depth range The acoustic emission anomalies and drilling parameter anomalies within the drilling system are analyzed, and the correlation coefficient between their characteristics is calculated. Based on the correlation results, the anomaly confidence level is calculated. .

5. The rock integrity prediction method based on drilling and CUSUM algorithm according to claim 3, characterized in that, The specific process for processing the video image data is as follows: Step T21: Construct a simulated borehole with a diameter of 93mm in the laboratory, with a target inside, collect target images from different angles, and establish a fisheye distortion model; Step T22: Extract lithological features from the image, distinguish lithology by the mean of grayscale histogram, and perform adaptive enhancement according to different lithologies; Step T23: Extract the standard deviation of the preprocessed image for lithology-adaptive Canny edge detection, perform morphological dilation on the initial edge image, connect the fracture edges, calculate the angle between the line segment and the borehole axis, and convert the angle difference... The line segments are classified as the same crack; Step T24: Record the drilling depth and image acquisition angle of the camera device, and establish the mapping relationship between image pixels and actual drilling positions; Step T25: Perform stereo matching on the crack edge and calculate the three-dimensional coordinates of the crack edge based on the principle of triangulation; Step T26: Calculate the orientation angle of the fracture line segment in the image, and correct the fracture direction by combining it with the actual tilt angle of the borehole axis through spatial geometric transformation.

6. The rock integrity prediction method based on drilling and CUSUM algorithm according to claim 5, characterized in that, In step T21, the fisheye distortion model adopts a polynomial distortion model: ;in The radius after distortion. For the ideal radius, , (where is the distortion coefficient); for borehole images acquired downhole, the ideal coordinates of each pixel are calculated based on a pre-established distortion model, and distortion correction is achieved through bilinear interpolation.

7. The rock integrity prediction method based on drilling and CUSUM algorithm according to claim 1, characterized in that, In step T26, the calculation process for crack width and orientation is as follows: When calculating the fracture width, the binocular camera device inside the borehole simultaneously acquires left and right images of the fracture area, performs adaptive Gamma correction and Retinex enhancement algorithm, calculates the three-dimensional coordinates of feature points in the left and right images based on the known intrinsic parameters of the binocular camera, and performs preliminary calculation of the fracture width; then, the vibration data corresponding to the fracture depth at the time is extracted, and wavelet packet decomposition is used to extract the energy in the 500Hz frequency band, establishes the relationship model between energy and fracture width, and obtains the final fracture width; When calculating the fracture orientation, an explosion-proof attitude sensor is installed at the end of the drill pipe of the TBM-mounted advanced drilling rig to collect the inclination angle and azimuth angle of the borehole in real time. The acquisition frequency is synchronized with the image inside the borehole to obtain the attitude parameters corresponding to the fracture depth. The fracture angle in the image plane is extracted, and a transformation model between the borehole local coordinate system and the downhole geodetic coordinate system is established. The fracture angle in the image plane is converted into the orientation angle in the geodetic coordinate system through Euler angle transformation, and finally the fracture orientation is determined.

8. The rock integrity prediction method based on drilling and CUSUM algorithm according to claim 1, characterized in that, In step S3, the specific process for aligning drilling parameters, acoustic emission events, and fracture images on the time-depth axis is as follows: Step S31: Acquire drilling parameters, acoustic emission events, and fracture images. Using the borehole start time as the time origin and the initial drill bit position as the depth origin, perform integral calculation of the depth using the drilling rate sensor. In the formula, for Drilling speed at any moment for Drilling depth at any given time; Step S32: Drilling parameters are directly mapped to the absolute time axis; acoustic emission events are time-stamped, and the trigger time of each event is recorded; fracture images are synchronously recorded with the current drilling depth based on the absolute timestamp of the capture time; Step S33: Use the modified formula Dynamic error compensation is performed, where, The elastic coefficient, for The thrust value of the drill pipe at any given time; Step S34: For the nth frame image, record its corresponding depth. , This refers to the fixed distance between the camera device and the drill bit; Step S35: Record the start time of the anomaly detected by the USUM algorithm. Corresponding depth ; Filter out time windows Acoustic emission events within the range of depth ; Retrieving depth in an image sequence The corresponding image frame is used to identify the crack features in that frame and the two adjacent frames.

9. The rock integrity prediction method based on drilling and CUSUM algorithm according to claim 1, characterized in that, In step S5, the crack density is calculated based on the detected crack conditions. And the rock mass quality indicators, the specific formulas are as follows: ;based on The values ​​are substituted into the formula to calculate the rock quality indicators respectively. and fracture volume factor The specific formula is as follows: ， 。 10. A rock integrity prediction method based on drilling and CUSUM algorithms according to claim 1, characterized in that, In step S6, the prediction results are fed back to the control console in the downhole operator's cab in real time. The main screen displays a three-dimensional rock mass integrity model. Based on the Kriging interpolation algorithm, different colors are used to render the rock mass quality within 50m in front of the tunnel face, with red representing the extremely fractured area and green representing the intact area. The secondary screen 1 displays real-time data curves, and the secondary screen 2 displays real-time images of the borehole.

Images