Deep hole drilling broken zone while drilling intelligent identification method and system

Abstract

The present application mainly relates to the field of drilling engineering technology, in order to solve the problem that the prediction of broken zone in deep hole drilling depends on ground analysis and the response speed is not enough, a method and system for intelligent identification of broken zone in deep hole drilling while drilling are provided, the core of which is to build a collaborative double-layer identification architecture on and under the well: the multi-source geological and engineering parameters are collected while drilling under the well, the characteristic values are extracted in real time through the built-in early warning model and the graded early warning signals are generated; the data transmission unit dynamically adjusts the transmission strategy according to the communication quality to ensure that the key information can be reliably delivered; after receiving the data, the ground unit rechecks the downhole early warning through correlation characteristic value mining and ground early warning model, predicts the broken zone risk probability and type, and finally generates accurate early warning decision, through the closed-loop mechanism of downhole real-time processing, hierarchical transmission guarantee and ground rechecking verification, the whole process intelligentization of broken zone in deep hole drilling from rapid response to accurate identification is realized, and the risk early warning capability of drilling engineering under complex working conditions is significantly improved.

Description

Technical Field

[0001] This invention relates to the field of drilling engineering technology, and in particular to a method and system for intelligent identification of fractured zones during deep hole drilling. Background Technology

[0002] Deep-hole drilling is a crucial technical means for mineral resource exploration, geological surveys, and underground engineering construction. As drilling depth increases, geological conditions become increasingly complex, especially with the presence of weak structural surfaces such as fractured zones. These conditions can easily lead to engineering accidents such as stuck drill bits, drill string burial, lost circulation, and wellbore collapse, severely impacting drilling efficiency and safety. Therefore, timely identification of fractured zones during drilling is of paramount importance for ensuring the smooth progress of the project.

[0003] Currently, monitoring while drilling (MSW) technology has been applied to some extent in drilling engineering. Existing methods typically involve installing sensors on the drill string to collect some physical parameters during the drilling process and transmitting the data to the ground for analysis. Ground technicians then use the changing trends of these parameters, combined with geological data and experience, to determine whether abnormal formations such as fracture zones have been encountered.

[0004] However, existing monitoring while drilling methods still have the following shortcomings in practical applications: Firstly, in terms of parameter acquisition, existing technologies are mostly limited to monitoring single types of parameters, making it difficult to comprehensively reflect the complex changes in the drilled strata. Different physical parameters have different responsiveness and characterization dimensions to strata properties, and a single parameter is often insufficient to accurately identify the boundaries, attitude, and filling characteristics of fracture zones.

[0005] Secondly, in terms of data processing, existing technologies mainly rely on centralized ground processing, requiring all raw data collected downhole to be transmitted to the surface before analysis can be performed. In deep hole drilling, as the drilling depth increases, the signal transmission distance becomes longer, channel attenuation intensifies, and data transmission rate is limited, resulting in poor real-time performance of data acquired from the ground, making it difficult to promptly detect and warn of sudden engineering risks.

[0006] Thirdly, in terms of information transmission, the data transmission channel between the well and the surface is often unstable due to factors such as drilling conditions, formation medium, and drilling technology. Existing transmission methods lack a flexible adjustment mechanism to cope with changes in communication quality. When communication conditions deteriorate, data delays and loss are likely to occur, affecting the continuity and reliability of monitoring.

[0007] Fourth, in terms of risk assessment, existing technologies largely rely on comprehensive analysis based on human experience, resulting in highly subjective judgment standards and making it difficult to achieve quantitative and standardized risk assessment. Different technical personnel may interpret the same monitoring data differently, leading to poor consistency in early warning results.

[0008] Therefore, how to improve the accuracy, real-time performance, and reliability of fracture zone identification during deep hole drilling is a technical problem that urgently needs to be solved in the field of drilling engineering technology. Summary of the Invention

[0009] The technical problem to be solved by this invention is to provide a method and system for intelligent identification of fractured zones in deep hole drilling, in order to solve the problem that the prediction of fractured zones in deep hole drilling relies on ground analysis and has insufficient response speed.

[0010] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows: On one hand, the present invention provides a method for intelligent identification of fractured zones during deep hole drilling, the method comprising: Real-time acquisition of multi-source geological and engineering parameters while drilling; The underground early warning model processes the collected parameters in real time, extracts the feature values ​​of each parameter, analyzes the feature values, and generates underground early warning signals. The downhole early warning signals and characteristic values ​​are transmitted to the surface through a hierarchical transmission method; Obtaining associated feature values ​​based on downhole early warning signals and feature values; The ground-based early warning model predicts the probability and type of risk in the fracture zone based on eigenvalues ​​and associated eigenvalues. The downhole early warning signal is reviewed based on the risk probability and type of the fractured zone to generate an early warning decision.

[0011] Furthermore, the geological parameters include at least one of electromagnetic wave reflection parameters, acoustic wave parameters, gamma ray parameters, and resistivity parameters; the engineering parameters include at least one of torque, pressure, and rotational speed parameters.

[0012] Furthermore, the electromagnetic wave reflection parameters include at least one of the reflected signal travel time, reflected signal amplitude, and reflected signal phase; the acoustic wave parameters include at least one of the acoustic wave dominant frequency, acoustic wave energy, acoustic wave propagation speed, and acoustic wave spectral characteristic parameters; the gamma ray parameter is the natural radioactivity count rate; and the resistivity parameters include formation resistivity and resistivity gradient.

[0013] Furthermore, the underground early warning model analyzes feature values ​​to generate downhole early warning signals by: judging the anomalies of geological and engineering parameters based on a set anomaly threshold, counting the number of abnormal features, calculating a weighted anomaly score according to a set weight, and generating early warning signals in a graded manner based on the weighted anomaly score.

[0014] Furthermore, the transmission of downhole early warning signals and characteristic values ​​to the surface via a hierarchical transmission method includes: When communication quality is good, priority is given to transmitting downhole early warning signals, followed by transmitting all characteristic values; When communication quality deteriorates, only downhole early warning signals and preset key characteristic values ​​are transmitted, and the data transmission frequency is increased; When communication is interrupted, the mine continuously sends early warning signals and records data locally. Once communication is restored, the signals are retransmitted.

[0015] Furthermore, the acquisition of associated feature values ​​based on downhole early warning signals and feature values ​​includes: Receive early warning signals uploaded from downhole, the early warning signals including the current early warning level and the corresponding set of feature values; Based on the preset rule base for associating warning levels with feature combinations, determine one or more feature combinations associated with the current warning level and the association weight of each feature. Feature values ​​corresponding to the feature combination are extracted from the feature value set, and coupled calculations are performed according to the association weights to generate associated feature values.

[0016] Furthermore, the underground early warning model is based on random forest, and the ground early warning model is based on XGBoost model.

[0017] On the other hand, the present invention also provides a deep hole drilling fracture zone intelligent identification system, the system comprising: a downhole integrated probe unit, a downhole early warning unit, a data transmission unit, a correlation feature value analysis unit, and a surface early warning unit; The downhole integrated probe unit is used to collect multi-source geological and engineering parameters while drilling; The downhole early warning unit is used to process the collected parameters in real time downhole, extract feature values, analyze the feature values ​​through the underground early warning model, and generate downhole early warning signals. The data transmission unit is used to upload downhole early warning signals and characteristic values ​​to the surface through a hierarchical transmission method; The correlation feature value analysis unit is used to obtain correlation feature values ​​based on downhole early warning signals and feature values; The ground-based early warning unit is used to predict the probability and type of risk in the fracture zone based on eigenvalues ​​and associated eigenvalues, to review the downhole early warning signals based on the probability and type of risk in the fracture zone, and to generate the final early warning decision.

[0018] Furthermore, the downhole integrated probe unit includes: an electromagnetic wave sensor, an acoustic wave sensor, a gamma ray sensor, a resistivity sensor, a torque sensor, a pressure sensor, and a speed sensor.

[0019] Furthermore, the system also includes a downhole data storage device for storing multi-source geological and engineering parameters, as well as generated early warning signals.

[0020] The beneficial effects of this invention are: (1) This invention achieves multi-dimensional perception of fractured zones by collecting various geological and engineering parameters while drilling, which improves the comprehensiveness and accuracy of identification. It establishes an early warning model downhole, extracts feature values ​​in real time and generates early warning signals, avoiding response lag caused by data transmission delay. It can issue early warnings as soon as abnormal formations are encountered. Furthermore, the ground early warning model verifies the downhole early warning signals based on the downhole early warning signals and combines them with associated feature values ​​to further assess the risk probability and type of fractured zones, forming a dual verification mechanism that significantly reduces false alarm rate and false alarm rate. (2) The data transmission strategy is dynamically adjusted according to the communication quality, and key early warning information is transmitted first to ensure that the early warning function can still be maintained in the event of poor or interrupted communication, and the data is retransmitted after recovery, thereby improving the robustness of the system. Attached Figure Description

[0021] Figure 1 A flowchart illustrating the overall workflow of the intelligent identification method for fractured zones during deep hole drilling. Figure 2 This is a flowchart of the underground early warning model. Figure 3 A flowchart for hierarchical data communication transmission; Figure 4 Flowchart for ground-based early warning review and decision-making process; Figure 5 Flowchart for training and inference of ground-based early warning model and ground-based risk assessment model; Figure 6 This is a schematic diagram of the structure of the downhole integrated exploration tube unit. Detailed Implementation

[0022] Due to technical problems such as delayed response, limited information, and unreliable transmission in existing technologies for identifying fractured zones under deep and complex geological conditions, this invention provides a method and system for predicting fractured zones in deep hole drilling. Its core lies in constructing a two-layer intelligent identification architecture that coordinates both uphole and ground-hole operations: the downhole unit collects multi-source geological and engineering parameters while drilling, extracts feature values ​​in real time through a built-in early warning model, and generates tiered early warning signals; the data transmission unit dynamically adjusts its transmission strategy based on communication quality to ensure reliable delivery of critical information; after receiving the data, the surface unit verifies the downhole early warning through correlation feature value mining and the surface early warning model, predicts the probability and type of fractured zone risk, and ultimately generates accurate early warning decisions. This invention, through a closed-loop mechanism of real-time downhole processing, tiered transmission assurance, and surface verification, achieves intelligent full-process identification of fractured zones in deep hole drilling, from rapid response to accurate identification, significantly improving the risk early warning capability of drilling projects under complex conditions.

[0023] The deep hole drilling fracture prediction method and system described in this invention will be further explained below with reference to the accompanying drawings and embodiments.

[0024] like Figure 1 As shown, the deep hole drilling fracture prediction method of the present invention includes the following steps: Step S1: Real-time acquisition of multi-source geological and engineering parameters while drilling.

[0025] Existing technologies rely on only one or two parameters, such as gamma rays combined with resistivity or acoustic waves, to identify complex geological bodies, which is insufficient. In reality, decreased resistivity may indicate water-rich plastic rock layers rather than fracture zones, and relying on a single parameter can easily lead to misjudgments. Therefore, this invention combines multiple geological parameters with engineering parameters and obtains characteristic values ​​of these parameters for downhole and surface early warning systems.

[0026] In this embodiment, the geological parameters include electromagnetic wave reflection parameters, acoustic wave parameters, gamma ray parameters, and resistivity parameters. The engineering parameters include torque, pressure, and rotational speed parameters.

[0027] The characteristic values ​​of the electromagnetic wave reflection parameters include: reflected signal travel time (ns): the time it takes for the electromagnetic wave to travel from transmission to reception, used to estimate the distance to the anomalous body; reflected signal amplitude (dB): the intensity of the reflected signal, indicating the magnitude of the electrical difference in the anomalous body; reflected signal phase (radians): the signal phase change, used to help determine the nature of the anomalous body; and estimated distance to the anomalous body ahead (m): a preliminary distance calculated based on travel time and wave velocity. The characteristic values ​​of the acoustic wave parameters include: acoustic wave dominant frequency (Hz): the dominant frequency of the rock-breaking acoustic signal from the drill bit, reflecting rock hardness; acoustic wave energy (dB): the total energy of the acoustic signal, indicating the degree of rock fragmentation; acoustic wave propagation speed (m / s): the wave speed from the transmitter to the receiver; sudden changes in wave speed may indicate changes in formation; and acoustic wave spectral characteristic parameters such as spectral width and peak frequency, used for lithology identification. The characteristic value of the gamma-ray parameter is the natural radioactivity count rate, which measures the natural radioactivity intensity of the formation and is used to distinguish lithology (e.g., shale is usually highly radioactive). The characteristic values ​​of the resistivity parameter include formation resistivity (Ω·m): identifying water-bearing or conductive fractures; resistivity gradient, the rate of change of resistivity with depth, and rapid changes may indicate fracture zones.

[0028] Torque characteristics include mean and variance (N·m): statistically indicating the stability of torque; abrupt changes may indicate encountering fractured zones. Pressure characteristics include mean and variance (MPa): variations in drilling fluid pressure, reflecting formation permeability. Rotational speed characteristics include mean and variance (rpm): stability of drill bit rotational speed; abnormal fluctuations may be associated with formation changes.

[0029] In this embodiment, the collection of multi-source geological parameters and engineering parameters specifically includes: like Figure 6As shown, sensors for collecting the aforementioned geological and engineering parameters are integrated onto the probe. Specifically: At the very front of the probe, closest to the drill bit, is an electromagnetic wave sensor. The sensor employs a coaxial arrangement of a transmitting coil (Tx) and a receiving coil (Rx). An electromagnetic shielding layer is placed between the Tx and Rx to prevent the transmitted signal from directly coupling to the receiving coil and drowning out the reflected signal. Electromagnetic waves propagate through the formation and are reflected when they encounter interfaces with electrical differences (such as the interface between intact rock strata and fractured zones). The receiving coil receives the reflected signal. By analyzing the travel time, amplitude, and phase of the reflected signal, the distance and size of anomalies ahead can be preliminarily determined, enabling advanced detection within a range of several meters. An acoustic wave sensor is installed behind the electromagnetic wave module. Multiple acoustic wave sensors are installed in an array. Their functions are twofold: first, to receive the acoustic / vibration signals generated by the drill bit breaking rock, analyze their spectral characteristics, and determine the mechanical properties of the rock; second, to receive the propagation speed of the acoustic waves generated by the acoustic wave transmitter in the formation, and sudden changes in wave velocity can indicate changes in the formation ahead. A gamma-ray sensor is installed behind the acoustic wave array to measure the natural radioactivity of the formation for lithological identification. A resistivity sensor is installed behind the acoustic array to measure the formation resistivity and identify water-bearing fracture zones.

[0030] In addition, torque sensors, pressure sensors, and speed sensors are installed on the probe. These sensors record engineering parameters such as downhole torque, pressure, and speed data in real time, and use data mutations to help determine whether a high-risk fracture zone has been encountered.

[0031] Step S2: The underground early warning model processes the collected parameters in real time, extracts the feature values ​​of each parameter, analyzes the feature values, and generates downhole early warning signals.

[0032] The underground early warning model adopts a quantized pruned random forest model, which is implemented as follows: Model selection and optimization: A quantized pruned random forest model is adopted, with a total number of model parameters ≤500KB and inference latency ≤100ms / inference (to ensure real-time response). Basic model training: Collect historical data from multiple typical wells, such as metal mine wells, oil and gas field wells, geothermal wells, and geological core wells, to train a basic random forest model and establish an underground early warning model; Working area fine-tuning: Before drilling in the current working area, the basic model is fine-tuned using the unlabeled data from the first 50 meters of drilling in the current working area. The model decision threshold is adjusted by calculating the characteristic value benchmarks of the intact rock strata in the working area, such as the average resistivity and average sonic velocity of intact granite, to ensure that it is adapted to the local geological conditions. The underground early warning model employs a multi-feature cross-validation mechanism to avoid misjudgments caused by a single feature anomaly. The early warning judgment logic is as follows: Figure 2 As shown, it specifically includes: 1) Real-time extraction of feature values ​​from multi-source geological and engineering parameters and input into the model; 2) The model performs anomaly detection for each feature. For example, a reflection amplitude ≥ 0.7 dB indicates an electromagnetic wave anomaly, and a resistivity ≤ 150 Ω•m indicates a resistivity anomaly. 3) Count the number and weight of abnormal features, and calculate the weighted outlier; 4) Classify early warning levels based on weighted outliers: Level 1 Warning (Alert): Weighted outlier is less than 0.3; Level 2 Warning: Weighted outlier greater than or equal to 0.3 and less than 0.6; Level 3 warning (high risk): Weighted outlier greater than or equal to 0.6 and less than 0.9; 5) Generate early warning signals: Use binary simplified encoding. The encoding rule is 00 to indicate no early warning, 01 to indicate a level 1 early warning, 10 to indicate a level 2 early warning, and 11 to indicate a level 3 early warning, minimizing the transmission bandwidth usage.

[0033] In another embodiment of the present invention, the underground early warning model can also be established based on a lightweight neural network model, and the basic model is fine-tuned using the current drilling area data, and early warning is carried out according to the above process.

[0034] Step S3: Transmit the downhole early warning signal and characteristic values ​​to the surface through a hierarchical transmission method.

[0035] High-speed mud pulse generator transmission technology or electromagnetic transmission while drilling is used to encode and upload downhole multi-sensor data to the surface.

[0036] To address the issue of large data transmission volumes for multiple parameters, this invention employs an adaptive encoding strategy. The core principle follows a hierarchical transmission logic: prioritizing downhole early warning signals, followed by feature values, and then offline extraction of complete data. This strategy adapts deep-hole communication quality to different scenarios, such as… Figure 3 As shown, it specifically includes: (1) In normal communication scenarios, such as when the signal-to-noise ratio is ≥15dB, the signal received from the ground is clear and there is little interference. At this time, the transmitted content is divided into two categories and is strictly ordered according to priority: First priority: Early warning signals generated downhole. As soon as an early warning is detected downhole, the signal will be transmitted immediately. If no early warning is detected, a no-warning signal will be transmitted every 2 seconds to ensure the surface can monitor the safety status in real time. Second priority: Complete characteristic values ​​of geological and engineering parameters. Transmit at a frequency of one set per second, and align data by the depth axis before transmission to avoid time differences caused by the transmission order of data from different sensors; In addition, CRC checksums are added to the transmitted warning signals and feature values. After the ground receives the data, it can automatically determine whether the data is complete. If a missing data is found, it will immediately request a retransmission from the mine to ensure data accuracy.

[0037] (2) In scenarios with poor communication quality, such as a signal-to-noise ratio of 10dB to 15dB, the ground-received signal is subject to interference and some data is easily lost. In order to avoid non-critical data occupying bandwidth and causing delays in the early warning signal, the transmitted content will be reduced, and only the core information will be retained. The data will still be transmitted according to priority. First priority: Downhole early warning signals. The transmission frequency is increased to retransmit once every 500 milliseconds to prevent missed connections on the surface due to signal interference; Second priority: Suspend the transmission of non-critical feature values ​​(such as gamma count rate and acoustic energy), prioritize the release of bandwidth for early warning signals and critical feature values, retain only the few types of feature values ​​most relevant to the judgment of the broken zone as critical feature values ​​for uploading, and reduce the transmission frequency of critical feature values. Use Huffman coding to further reduce the number of transmitted bytes and reduce the impact of interference. Resend the critical feature values ​​during the suspension period after the communication quality is restored.

[0038] (3) Communication interruption / extreme conditions, such as high sand content in drilling fluid, signal attenuation in deep holes over 3000 meters resulting in a signal-to-noise ratio of <10dB, and the ground basically cannot receive effective signals, adopt pulse wake-up transmission mode, actively send a warning signal once every 5 seconds (regardless of whether there is a warning), even if the ground cannot receive it temporarily, it will continue to send until communication is restored, and the downhole will automatically record all warning records during the interruption period, including warning time, level, and corresponding characteristic value. After communication is restored, these interruption records will be retransmitted first, and then the transmission will be restored according to the normal scenario; at the same time, the power consumption of the mud pulse generator will be reduced to avoid power failure of downhole equipment due to continuous high load.

[0039] S4: Obtain associated feature values ​​based on downhole early warning signals and feature values.

[0040] Historical drilling data is collected, and the geological types and actual warning levels of each depth segment are labeled. For each warning level, the distribution differences of each feature value are statistically analyzed. Features that are significantly related to the level are screened out through mutual information or chi-square test. According to the preset warning level and feature combination association rule library, one or more feature combinations associated with the current warning level and the association weight of each feature are determined. Feature values ​​corresponding to the feature combinations are extracted from the feature value set, and coupled calculations are performed according to the association weights to generate associated feature values.

[0041] Step S5: The ground-based early warning model predicts the probability and type of risk in the fracture zone based on eigenvalues ​​and associated eigenvalues; like Figure 4 and Figure 5As shown, the ground-based early warning model uses the XGBoost gradient boosting tree model, and the training process is as follows: Data collection and labeling: Historical fracture zone identification data were collected, and geological experts labeled the fracture zones with three categories: non-fractured zones, dry fractured zones, and water-bearing fractured zones.

[0042] The received feature values ​​are processed as follows: data alignment, aligning feature values ​​from different sensors according to the depth axis; denoising, using Kalman filtering to remove random noise from the feature values; standardization, using the mean of feature values ​​of intact rock formations in the current working area as a benchmark, normalizing the feature values, and forming a multi-dimensional feature vector based on the feature values ​​extracted from downhole and associated feature values; Model training: Five-fold cross-validation was used, and hyperparameters were adjusted (e.g., tree depth = 8, learning rate = 0.1, estimators = 100) to improve the model's accuracy and recall. Model deployment: Export the trained model to ONNX format and integrate it into ground software to support real-time inference.

[0043] The surface early warning model prioritizes parsing downhole early warning signals, especially 1-byte signals. Upon parsing, a pop-up notification appears on the display screen, triggering an audible and visual alarm. Subsequently, the downhole early warning verification logic is based on the predicted probability, confidence level, and type of the fracture zone risk according to eigenvalues ​​and associated eigenvalues. The specific process is as follows: When a Level 1 warning (notification) is issued downhole, if the risk probability in the surface model is ≥30%, the Level 1 warning is maintained; if it is <30%, it is downgraded to "no warning" and marked as "downhole misjudgment" in the software. When a Level 2 warning (alert) is issued downhole, if the risk probability in the surface model is ≥60%, the Level 2 warning is maintained; if the probability is 30% ≤ < 60%, it is downgraded to a Level 1 warning; if the probability is < 30%, the warning is cancelled. When a Level 3 downhole warning (high risk) is triggered, if the risk probability in the surface model is ≥85%, the Level 3 warning is maintained (and an emergency drilling stop recommendation is triggered); if the probability is ≤60% and <85%, it is downgraded to a Level 2 warning; if the probability is <60%, the data is further verified (e.g., whether there is a sensor malfunction). After obtaining the verification results, the downhole warning level, the surface verification level, and the risk probability are displayed on the screen simultaneously, and marked as double confirmation (downhole and surface levels are consistent) or level adjustment (downhole and surface levels are inconsistent) to facilitate operator decision-making.

[0044] S6: Based on the risk probability and type of the fractured zone, review the downhole early warning signal, generate an early warning decision, and handle the risk according to the early warning decision. Specific rules: Level 1 Risk Handling: If the probability of risk during ground verification reaches 30%, or if a weak anomaly (reflection amplitude 0.3-0.7 dB) is detected 5-10 meters ahead via electromagnetic wave detection, the operator should be advised to "pay attention to changes in drilling pressure and torque, and prepare to adjust drilling parameters." Level 2 Risk Handling: The probability of risk during ground verification reaches 60%, and current engineering parameters (such as torque variance) fluctuate (fluctuation amplitude ≥ 50%). Warn the operator to "immediately reduce drilling pressure (to 70% of the current value), reduce drilling speed (to 80% of the current value), and prepare high-viscosity drilling fluid in advance." Level 3 Risk Handling: Ground verification risk probability > 85%, and engineering parameters deteriorate drastically (torque fluctuation ≥ 100%, pump pressure change ≥ 2 MPa). The system immediately triggers a red bell alarm, displays a suggestion to "immediately stop the advance and initiate wall protection measures (such as increasing mud viscosity to 50s and shear force to 15Pa)," and records the current depth and alarm time for subsequent traceability.

[0045] The deep hole drilling fracture zone intelligent identification system of the present invention includes: a downhole integrated probe unit, a downhole early warning unit, a data transmission unit, a correlation feature value analysis unit, and a surface early warning unit; The downhole integrated probe unit is used to collect multi-source geological and engineering parameters while drilling; the downhole integrated probe unit includes: electromagnetic wave sensor, acoustic wave sensor, gamma ray sensor, resistivity sensor, torque sensor, pressure sensor and speed sensor; The downhole early warning unit is used to process the collected parameters in real time downhole, extract feature values, analyze the feature values ​​through the underground early warning model, and generate downhole early warning signals. The data transmission unit is used to upload downhole early warning signals and characteristic values ​​to the surface through a hierarchical transmission method; The correlation feature value analysis unit is used to obtain correlation feature values ​​based on downhole early warning signals and feature values; The ground-based early warning unit is used to predict the probability and type of risk in the fracture zone based on eigenvalues ​​and associated eigenvalues, to review the downhole early warning signals based on the probability and type of risk in the fracture zone, and to generate the final early warning decision.

[0046] Furthermore, the system also includes a downhole data storage device for storing multi-source geological and engineering parameters, as well as generated early warning signals.

Claims

1. A method for intelligent identification of fractured zones during deep hole drilling, characterized in that, The method includes: Real-time acquisition of multi-source geological and engineering parameters while drilling; The underground early warning model processes the collected parameters in real time, extracts the feature values ​​of each parameter, analyzes the feature values, and generates underground early warning signals. The downhole early warning signals and characteristic values ​​are transmitted to the surface through a hierarchical transmission method; Obtaining associated feature values ​​based on downhole early warning signals and feature values; The ground-based early warning model predicts the probability and type of risk in the fracture zone based on eigenvalues ​​and associated eigenvalues. The downhole early warning signal is reviewed based on the risk probability and type of the fractured zone, an early warning decision is generated, and risk processing is carried out according to the early warning decision.

2. The intelligent identification method for fractured zones during deep hole drilling according to claim 1, characterized in that, The geological parameters include electromagnetic wave reflection parameters, acoustic wave parameters, gamma ray parameters, and resistivity parameters; the engineering parameters include torque, pressure, and rotational speed parameters.

3. The intelligent identification method for fractured zones in deep hole drilling according to claim 2, characterized in that, The electromagnetic wave reflection parameter features include the travel time of the reflected signal, the amplitude of the reflected signal, and the phase of the reflected signal; the acoustic wave parameter features include the dominant frequency of the acoustic wave, the energy of the acoustic wave, the propagation speed of the acoustic wave, and the spectral features of the acoustic wave; the gamma ray parameter features are the natural radioactivity count rate; and the resistivity parameter features include the formation resistivity and the resistivity gradient.

4. The intelligent identification method for fractured zones in deep hole drilling according to claim 1, characterized in that, The underground early warning model analyzes feature values ​​and generates downhole early warning signals by: judging the anomalies of geological and engineering parameters based on set anomaly thresholds, counting the number of abnormal features, calculating weighted anomaly scores according to set weights, and generating early warning signals in a graded manner based on the weighted anomaly scores.

5. The intelligent identification method for fractured zones during deep hole drilling according to claim 1, characterized in that, The transmission of downhole early warning signals and characteristic values ​​to the surface via a hierarchical transmission method includes: When communication quality is good, priority is given to transmitting downhole early warning signals, followed by transmitting all characteristic values; When communication quality deteriorates, only downhole early warning signals and preset key characteristic values ​​are transmitted, and the data transmission frequency is increased; When communication is interrupted, the mine continuously sends early warning signals and records data locally. Once communication is restored, the signals are retransmitted.

6. The intelligent identification method for fractured zones in deep hole drilling according to claim 1, characterized in that, The method of obtaining associated feature values ​​based on downhole early warning signals and feature values ​​includes: Receive early warning signals uploaded from downhole, the early warning signals including the current early warning level and the corresponding set of feature values; Based on the preset rule base for associating warning levels with feature combinations, determine one or more feature combinations associated with the current warning level and the association weight of each feature. Feature values ​​corresponding to the feature combination are extracted from the feature value set, and coupled calculations are performed according to the association weights to generate associated feature values.

7. The intelligent identification method for fractured zones in deep hole drilling according to claim 1, characterized in that, The underground early warning model is based on random forest or neural network, while the ground early warning model is based on XGBoost model.

8. A deep-hole drilling fracture zone intelligent identification system for implementing the deep-hole drilling fracture zone intelligent identification method according to any one of claims 1-7, characterized in that, The system includes: a downhole integrated pipe exploration unit, a downhole early warning unit, a data transmission unit, a correlation feature value analysis unit, and a surface early warning unit; The downhole integrated probe unit is used to collect multi-source geological and engineering parameters while drilling; The downhole early warning unit is used to process the collected parameters in real time downhole, extract feature values, analyze the feature values ​​through the underground early warning model, and generate downhole early warning signals. The data transmission unit is used to upload downhole early warning signals and characteristic values ​​to the surface through a hierarchical transmission method; The correlation feature value analysis unit is used to obtain correlation feature values ​​based on downhole early warning signals and feature values; The ground-based early warning unit is used to predict the probability and type of risk in the fracture zone based on eigenvalues ​​and associated eigenvalues, to review the downhole early warning signals based on the probability and type of risk in the fracture zone, and to generate the final early warning decision.

9. The intelligent identification system for fractured zones in deep hole drilling according to claim 8, characterized in that, The downhole integrated probe unit includes: an electromagnetic wave sensor, an acoustic wave sensor, a gamma ray sensor, a resistivity sensor, a torque sensor, a pressure sensor, and a speed sensor.

10. The intelligent identification system for fractured zones in deep hole drilling according to claim 8, characterized in that, The system also includes a downhole data storage device for storing multi-source geological and engineering parameters, as well as generated early warning signals.

Images