Method for determining different damage modes of pneumatic percussion based on drilling progress monitoring

Abstract

The present application relates to the technical field of drilling process monitoring, and particularly relates to a method for determining pneumatic impact different damage modes based on drilling progress monitoring. The method comprises the following steps: obtaining time schedule monitoring data of a drilling process; extracting a net drilling process by using an improved window energy calculation model to obtain net drilling process data; extracting first drilling feature parameters and second drilling feature parameters; determining a preliminary damage mode recognition result by using a first judgment condition; and determining an optimized damage mode recognition result by using a second judgment condition. The present application extracts a net drilling process by using an improved window energy calculation model, combines impact pressure fluctuation intensity and displacement mutation intensity parameters, constructs a damage mode recognition model, can accurately distinguish between crushing mode and cutting mode, improves result reliability through an optimized recognition process, provides data support for drilling parameter optimization, hole collapse early warning and wall protection selection, and effectively improves hole quality and drilling efficiency.

Description

Technical Field

[0001] This invention relates to the field of drilling process monitoring technology, specifically to a method for determining different failure modes of aerodynamic impact based on drilling progress monitoring. Background Technology

[0002] Drilling process monitoring technology is widely used in geotechnical engineering and pile foundation construction for borehole quality assessment and geological condition analysis. In recent years, with the development of microprocessor and sensor technology, drilling process monitoring systems can collect and process multi-dimensional time-series data such as impact pressure, rotational pressure, and drill bit displacement in real time, providing a data foundation for identifying drilling status and rock mass response.

[0003] Existing methods are mostly based on simple time-domain statistical analysis or fixed threshold judgments. For example, they judge drilling anomalies by the mean or variance changes of impact pressure, or divide the process into stages by combining displacement rate. These methods have initially achieved the identification of typical problems such as stuck drill and hole collapse. However, the following shortcomings still exist: First, the extraction of the net drilling process relies heavily on manual experience or simple energy thresholds, making it difficult to effectively eliminate invalid periods such as idling and drill lifting, resulting in inaccurate feature extraction. Second, the feature parameter design is relatively simple and fails to fully integrate the characteristics of impact pressure fluctuations and displacement mutation behavior, resulting in limited ability to distinguish different failure modes. Third, in the pattern recognition stage, there is a lack of robust processing mechanisms for feature overlap and noise interference, making the recognition results susceptible to fluctuations in operating conditions and lacking reliability.

[0004] To address the aforementioned problems, this invention proposes a method for determining different failure modes of aerodynamic impact based on borehole progress monitoring. This method achieves accurate extraction of the net drilling process through an improved window energy model, uses the impact pressure fluctuation intensity and displacement mutation intensity as key features, and introduces confidence assessment and graph neural network optimization mechanisms to achieve high-precision and adaptive identification of typical failure modes such as crushing mode and cutting mode, providing a reliable basis for drilling parameter optimization and borehole quality control. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for determining different failure modes of aerodynamic impact based on borehole progress monitoring, comprising the following steps: acquiring time-history monitoring data of the drilling process, wherein the time-history monitoring data includes time-series data of impact pressure, rotation pressure, and drill bit position; extracting net drilling process data using an improved window energy calculation model based on the time-history monitoring data; extracting a first drilling characteristic parameter and a second drilling characteristic parameter using the net drilling process data; determining a preliminary failure mode identification result based on the first and second drilling characteristic parameters and a first judgment condition; determining an optimized failure mode identification result based on the preliminary failure mode identification result and a second judgment condition; and determining different failure modes of aerodynamic impact using the optimized failure mode identification result.

[0006] Optionally, based on the time-history monitoring data, the net drilling process is extracted using an improved window energy calculation model to obtain net drilling process data, including: constructing an improved window energy calculation model by introducing a weighted power spectral density integral and a transient impact event enhancement mechanism based on the time-history monitoring data; obtaining a window energy value using the improved window energy calculation model; determining a high-energy drilling sampling window using an adaptive dual-threshold method based on the window energy value; constructing a drill bit displacement change rate calculation model by introducing a sliding fitting slope and an acceleration suppression mechanism based on the high-energy drilling sampling window; calculating the drill bit displacement change rate using the drill bit displacement change rate calculation model; establishing a displacement change rate threshold determination model based on drilling equipment performance, estimated rock mass strength, and historical drilling data; calculating the displacement change rate threshold using the displacement change rate threshold determination model; comparing the drill bit displacement change rate with the displacement change rate threshold, and determining a net drilling process if the drill bit displacement change rate is consistently higher than the displacement change rate threshold; and extracting time-history monitoring data within the time period based on the net drilling process to obtain net drilling process data.

[0007] Optionally, the improved window energy calculation model satisfies the following expression:

[0008] ,

[0009] in, For a moment The enhanced energy value of the window in question. This represents the number of sampling points within the sliding window. For the first in the window Impact pressure values ​​at each sampling point For frequency domain energy weighting coefficients, For frequency The weight function, For a moment Power spectral density of nearby signals, , These are the lower and upper limits of the characteristic frequency band of the effective drilling signal, respectively. For the Nyquist frequency, For transient impact weighting coefficients, This is a set of indices for transient impact events detected within the window. For the first The amplitude of the impact pressure change in a transient impact event.

[0010] Optionally, extracting the first and second drilling characteristic parameters using the net drilling process data includes: establishing an impact pressure fluctuation intensity calculation model using the net drilling process data; calculating the first drilling characteristic parameter, which includes the impact pressure fluctuation intensity, using the impact pressure fluctuation intensity calculation model; constructing a displacement mutation intensity calculation model using the net drilling process data; calculating the second drilling characteristic parameter, which includes the displacement mutation intensity, using the displacement mutation intensity calculation model; and standardizing the first and second drilling characteristic parameters to obtain standardized first and second drilling characteristic parameters.

[0011] Optionally, the impact pressure fluctuation intensity calculation model satisfies the following relationship:

[0012] ,

[0013] in, The intensity of the impact pressure fluctuation. For the first Impact pressure values ​​at each sampling point To calculate the mean impact pressure within the window, The length of the sliding window. The range impact factor, , These represent the maximum and minimum values ​​of the impact pressure within the window, respectively. This is the pressure normalization coefficient. The instantaneous change weighting coefficients, and the displacement mutation intensity calculation model satisfy the following expression:

[0014] ,

[0015] in, The intensity of the sudden displacement. for The drill bit position value at any given time. The sampling time interval; To calculate the number of sampling points within the window, This represents the mean of the drill bit position sequence within the window. , These are the mean and variance of the drill bit position sequence throughout the entire net drilling process, respectively.

[0016] Optionally, determining the preliminary damage mode identification result based on the first drilling feature parameters and the second drilling feature parameters, and through a first judgment condition, includes: establishing a calculation model for the confidence level of the breaking mode and the confidence level of the cutting mode based on the first drilling feature parameters and the second drilling feature parameters; calculating the confidence level of the breaking mode and the confidence level of the cutting mode using the calculation model; calculating the confidence level difference ratio based on the confidence level of the breaking mode and the confidence level of the cutting mode; determining a first preset threshold; comparing the confidence level difference ratio with the first preset threshold; if the confidence level difference ratio is greater than the first preset threshold, then determining the mode with high confidence as the preliminary damage mode identification result; if the confidence level difference ratio is not greater than the first preset threshold, then triggering a result arbitration mechanism based on feature clustering, and determining the mode to which the cluster center belongs as the preliminary damage mode identification result.

[0017] Optionally, based on the preliminary damage pattern identification result, determining the optimized damage pattern identification result through a second judgment condition includes: establishing a location uncertainty quantification value calculation model based on the preliminary damage pattern identification result; determining the location uncertainty quantification value through the location uncertainty quantification value calculation model; establishing a time series continuity index calculation model based on the preliminary damage pattern identification result; determining the time series continuity index through the time series continuity index calculation model; and determining a second preset threshold.

[0018] The location uncertainty quantization value is compared with the second preset threshold. If the location uncertainty quantization value is not greater than the second preset threshold, the preliminary damage pattern recognition result is used as the optimized damage pattern recognition result. If the location uncertainty quantization value is greater than the second preset threshold, the result optimization process based on the second judgment condition is triggered. The optimized damage pattern recognition result is determined according to the comparison result between the location uncertainty quantization value and the second preset threshold.

[0019] Optionally, the result optimization process based on the second judgment condition includes: calculating the consistency score between drilling feature parameters and the historical typical pattern database; determining a third preset threshold; comparing the consistency score with the third preset threshold: if the consistency score is not less than the third preset threshold, then the preliminary damage pattern recognition result is used as the optimized damage pattern recognition result; if the consistency score is less than the third preset threshold, then feature reconstruction and pattern re-identification based on graph neural networks are initiated, and the re-identification result is used as the optimized damage pattern recognition result.

[0020] Optionally, the feature reconstruction and pattern re-identification based on graph neural networks includes: constructing a feature heterogeneous graph; using a graph attention network to aggregate and enhance node features of the feature heterogeneous graph to obtain a reconstructed high-dimensional feature vector; inputting the high-dimensional feature vector into a pre-trained pattern classifier to obtain the probability of re-identified patterns; and based on the probability of re-identified patterns, performing final arbitration in conjunction with an expert rule base to output an optimized destructive pattern identification result, wherein the optimized destructive pattern identification result includes pattern type, confidence level, optimization flag, and suggestion confidence level.

[0021] Optionally, determining different aerodynamic impact failure modes using the optimized failure mode identification results includes: obtaining a drilling parameter optimization strategy library using the optimized failure mode identification results; establishing an adaptive weight allocation model based on the drilling parameter optimization strategy library; calculating recommended drilling parameters using the adaptive weight allocation model, the recommended drilling parameters including recommended thrust and recommended slewing speed; constructing a drilling status report based on the recommended drilling parameters, the drilling status report including optimized failure mode identification results, recommended drilling parameters, real-time risk assessment level, and wall protection suggestions; and determining different aerodynamic impact failure modes using the drilling status report.

[0022] Compared with the prior art, the beneficial effects of the present invention include:

[0023] 1. Deep fusion and high-precision synchronous acquisition of multi-source high-frequency monitoring data were achieved; by deploying multiple types of sensors and combining GPS timing with a crystal oscillator complementary synchronous clock module, the time alignment accuracy of multi-channel data was better than 1 millisecond; an industrial-grade embedded system was used for multi-channel synchronous sampling and real-time storage, ensuring data integrity and timeliness; and the problems of data asynchrony, low sampling rate, and large noise interference in existing methods were solved.

[0024] 2. An improved window energy calculation model was proposed, which enabled high-precision automatic extraction during the net drilling process;

[0025] Based on the existing root mean square energy, a weighted power spectral density integral and transient impact event enhancement mechanism are introduced to improve the sensitivity to specific frequency components in non-stationary signals. Combined with the adaptive double threshold method and secondary confirmation of displacement change rate, invalid periods such as idling and drilling are effectively eliminated. The accuracy and robustness of the net drilling process extraction are significantly improved.

[0026] 3. A multi-dimensional and highly sensitive drilling characteristic parameter system was constructed; two key characteristic parameters, impact pressure fluctuation intensity and displacement mutation intensity, were proposed to quantify rock mass failure characteristics from the two dimensions of dynamic pressure change and displacement discontinuity, respectively; by introducing mechanisms such as range influence factor, instantaneous change weight, and global normalization, the ability of features to distinguish failure modes was enhanced; the shortcomings of existing methods in representing complex failure modes were overcome, and the sensitivity and stability of pattern recognition were improved.

[0027] 4. A pattern recognition system based on confidence assessment and arbitration mechanism was established; the Mahalanobis distance weighted probability mapping method was used to calculate the pattern confidence, and the confidence difference ratio and cluster arbitration mechanism were combined to achieve preliminary pattern recognition; the quantitative analysis of positional uncertainty and the temporal continuity index were introduced to refine the preliminary results; when the uncertainty is high, graph neural network feature reconstruction and re-recognition were initiated to improve the robustness of recognition under complex working conditions; the accuracy and reliability of destructive pattern recognition were significantly improved, especially in complex scenarios such as feature overlap and noise interference.

[0028] 5. A complete closed-loop system from identification to decision-making has been formed; based on the identification results, combined with historical data and expert knowledge, recommended drilling parameters, including thrust and rotation speed, are dynamically generated; a structured report containing mode type, risk assessment and wall protection suggestions is output, supporting manual intervention or automatic control; the entire process from perception to identification to decision-making to control has been made intelligent, significantly improving drilling efficiency, hole quality and operational safety.

[0029] 6. It possesses excellent engineering verification and application value; through on-site verification methods such as grouting tests, it has been confirmed that the identification results are highly consistent with the actual situation; the system has good scalability and can be adapted to different types of drilling rigs and rock conditions; it provides reliable technical support for engineering projects such as slope protection, pile foundation drilling, and mining, and has broad engineering application prospects. Attached Figure Description

[0030] Figure 1 This is a flowchart of a method for determining different failure modes of aerodynamic impact based on borehole progress monitoring, according to an embodiment of the present invention.

[0031] Figure 2 This is a schematic diagram of the monitoring data of the pneumatic impact drilling rig throughout the drilling process according to an embodiment of the present invention;

[0032] Figure 3 This is a schematic diagram of the net drilling process of the pneumatic impact drill according to an embodiment of the present invention;

[0033] Figure 4 The graphs showing the changes in drill bit position, downward pressure, and impact pressure over time during the borehole drilling process, as described in this embodiment of the invention, are examples of the present invention. Detailed Implementation

[0034] Specific embodiments of the present invention will now be described in detail. It should be noted that the embodiments described herein are for illustrative purposes only and are not intended to limit the invention. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that these specific details are not necessary to practice the invention. In other instances, well-known circuits, software, or methods have not been specifically described to avoid obscuring the invention.

[0035] Throughout this specification, references to "an embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with that embodiment or example is included in at least one embodiment of the invention. Therefore, the phrases "in an embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily refer to the same embodiment or example. Furthermore, specific features, structures, or characteristics can be combined in one or more embodiments or examples in any suitable combination and / or sub-combination. Moreover, those skilled in the art will understand that the illustrations provided herein are for illustrative purposes and are not necessarily drawn to scale.

[0036] Please see Figure 1 The embodiments of the present invention provide a method for determining different failure modes of aerodynamic impact based on drilling progress monitoring, the method comprising the following steps:

[0037] S1. Obtain time-series monitoring data of the drilling process, wherein the time-series monitoring data includes time-series data of impact pressure, rotation pressure and drill bit position.

[0038] In one embodiment, a drilling process monitoring system is first deployed. The system includes multiple types of sensors, a data acquisition unit, and a synchronization clock module installed on a pneumatic impact drilling rig. The sensors include an impact pressure sensor, a rotary pressure sensor, and a drill bit displacement sensor. The data acquisition unit adopts an industrial-grade embedded system with multi-channel synchronous sampling and real-time storage capabilities. The synchronization clock module uses GPS timing and crystal oscillator complementarity to ensure that the time alignment accuracy of various monitoring data is better than 1 millisecond. Before the drilling operation starts, the monitoring system performs system self-test and sensor calibration, including static pressure calibration of the impact pressure sensor, linearity verification of the rotary pressure sensor, and zero-point calibration of the drill bit displacement sensor, and records the calibration parameters for subsequent data correction.

[0039] Furthermore, time-series data of the impact pressure were obtained.

[0040] Specifically, the impact pressure sensor is installed near the drill bit end of the drilling rig's impact cylinder. It is a piezoresistive pressure sensor with a range covering 1.5 times the rated impact pressure of the drilling rig. The sampling frequency is set to 1000 Hz to capture high-frequency pressure fluctuations during the impact process. During the acquisition process, the sensor's working status is monitored in real time, and high-frequency electrical noise is removed through a hardware filtering circuit. The acquired raw impact pressure signal is preprocessed, including moving average filtering to remove high-frequency interference, outlier removal using outlier detection based on the Laida criterion, and signal drift compensation using linear fitting correction. Finally, clean impact pressure time-series data is obtained, and a timestamp and sensor identification are embedded.

[0041] Furthermore, the timing data of the slewing pressure is obtained.

[0042] Specifically, the rotary pressure sensor is installed in the oil inlet circuit of the drilling rig's rotary motor. It is a strain gauge type pressure sensor with a sampling frequency set to 500 Hz to adapt to the mid-frequency dynamic characteristics of the rotary system. During the acquisition process, hydraulic oil temperature data is recorded simultaneously for temperature-based compensation and correction of the pressure sensing characteristics. The raw rotary pressure signal undergoes preprocessing, including adaptive denoising based on wavelet thresholds, polynomial fitting for trend term removal, and phase-preserving low-pass filtering for pressure pulsation smoothing. This results in high-quality rotary pressure time-series data, which is then associated with corresponding timestamps and operating condition markers.

[0043] Furthermore, the drilling position time series data of the drill bit is obtained.

[0044] Specifically, the drill bit displacement sensor is a magnetostrictive linear displacement sensor, installed on the drilling rig's propulsion mechanism, directly measuring the absolute displacement of the drill bit relative to the drill frame with a measurement accuracy of ±0.1 mm and a sampling frequency of 200 Hz. During data acquisition, the drilling rig's attitude data, such as inclination and azimuth, are monitored in real time, and the raw displacement data is converted into a displacement perpendicular to the drilling face through coordinate transformation. The raw displacement signal is preprocessed, including outlier removal using median filtering, data interpolation using cubic spline interpolation to fill missing points, and drill rod elastic deformation compensation based on a material mechanics model for reverse correction. Finally, accurate drill bit drilling position time-series data is obtained, and timestamps and spatial coordinate information are recorded simultaneously.

[0045] Furthermore, the synchronization and integration of time-series monitoring data.

[0046] Specifically, the data acquisition unit achieves multi-channel synchronous sampling through hardware triggering, and the data streams from all sensors are aligned using a unified time reference. After acquisition, the impact pressure time-series data, slewing pressure time-series data, and drill bit position time-series data are merged according to time series to form a structured time-history monitoring dataset. The dataset includes timestamps, impact pressure values, slewing pressure values, drill bit position values, and corresponding data quality flags, and is stored in a standard time-series format for subsequent analysis modules to access.

[0047] like Figure 2 The diagram shows the monitoring data of the entire drilling process of the pneumatic impact drilling rig. It displays the original curves of impact pressure, rotation pressure and drill bit position changing over time, which are collected synchronously from multiple channels, providing a data basis for subsequent net drilling process extraction.

[0048] S2. Based on the time history monitoring data, the net drilling process is extracted using an improved window energy calculation model to obtain net drilling process data.

[0049] In one embodiment, based on the time-history monitoring data obtained in step S1, net drilling process data representing effective drilling operations is extracted through a multi-stage data filtering and logical fusion process. The extraction of net drilling process aims to eliminate invalid periods in drilling operations, such as drill bit lifting, rod changing, and idling, while retaining efficient drilling periods in which the drill bit and rock mass continuously interact, thus providing a high-quality data foundation for subsequent feature extraction and pattern recognition.

[0050] Specifically, firstly, a sliding window energy detection is performed on the time-history monitoring data to identify high-energy drilling sampling windows. The sliding window energy detection employs an adaptive threshold determination method based on the short-time energy of the impact pressure signal. The key focus is on constructing an improved window energy calculation model. This model introduces a weighted power spectral density integral and a transient impact event enhancement mechanism on top of the existing root-mean-square energy model to more accurately characterize the dynamic energy distribution of the drilling process. The improved window energy calculation model satisfies the following expression:

[0051] ,

[0052] in, For a moment The enhanced energy value of the window is used to quantify the intensity of drilling activity during that period; The number of sampling points within the sliding window is determined by both the window length and the sampling frequency. For the first in the window The impact pressure values ​​at each sampling point are directly obtained from the impact pressure time series data after preprocessing in step S1. The frequency domain energy weighting coefficient is used to adjust the contribution of the power spectral density integral term to the total energy. It is obtained through regression analysis of drilling experimental data of different rock strata. For frequency The weighting function is designed based on the frequency band characteristics of typical effective drilling signals, and adopts the form of raised cosine function to highlight the effective drilling frequency band. For a moment The power spectral density of nearby signals was calculated using the windowed periodogram method. , These are the lower and upper limits of the characteristic frequency band of the effective drilling signal, determined based on the drilling rig-rock mass interaction mechanism and experimental spectrum analysis. The Nyquist frequency is determined by the sampling frequency; This is the transient impact weighting coefficient, used to increase the proportion of sudden high-voltage pulse events in energy calculation. It is obtained by optimizing the impact event identification accuracy. The set of indexes of transient impact events detected within the window is obtained by first-order differential over-threshold detection of the impact pressure signal; For the first The magnitude of the impact pressure change in a transient impact event is calculated using the pressure difference before and after the event.

[0053] Compared with existing technologies, the improvements of the improved window energy calculation model are: overcoming the shortcomings of existing methods in being insensitive to specific frequency components in non-stationary signals, such as the characteristic frequency of rock mass fracture; avoiding continuous stationary signals, such as idling, from being misjudged as high-energy drilling, and improving the ability to identify effective drilling windows.

[0054] Furthermore, based on the improved window energy calculation model, a sliding calculation is performed on the entire time-series monitoring data to obtain the window energy time series sequence; subsequently, an adaptive double threshold method is used to determine the high-energy window: specifically, the upper quartile of the window energy sequence is taken as the high threshold. This is used to screen for confirmed high-energy drilling periods; the median of the window energy sequence is taken as the low threshold. It is also fine-tuned in conjunction with the background noise level to capture potential low-energy effective drilling.

[0055] If window energy If so, it is directly marked as a high-energy drilling sampling window;

[0056] If window energy If so, it is directly marked as a low-energy drilling sampling window;

[0057] like Then, a second confirmation is required based on the subsequent displacement change rate; finally, a list of start and end times for the high-energy drilling sampling window is output.

[0058] Furthermore, based on the high-energy drilling sampling window, a drill bit displacement change rate calculation model is constructed, and the time-series data of displacement change rate are calculated using this model. This model introduces a sliding fitting slope and acceleration suppression mechanism on the basis of the standard first-order difference to smooth noise interference and more stably reflect the true drilling trend of the drill bit. The drill bit displacement change rate calculation model satisfies the following expression:

[0059] ,

[0060] in, For a moment The rate of change of drill bit displacement, i.e., drilling speed, characterizes the amount of forward movement of the drill bit per unit time. The half-width of the sliding fitting window is set according to the inertial characteristics of the drill bit's motion; For the first The weighting coefficients for each sampling point are allocated using a Gaussian window function to enhance the contribution of the center point; For a moment The drill bit position value is obtained from the drill bit position time series data after preprocessing in step S1; , These are the mean values ​​of the location and time within the fitting window, respectively; This is an acceleration suppression factor used to reduce the impact of severe acceleration fluctuations on velocity estimation. It is obtained through optimization of the mean square error of velocity estimation. For a moment The estimated value of the drill bit acceleration is obtained by further differencing the displacement rate sequence. The reference acceleration value is used for normalization, and the maximum acceleration statistical value during a typical drilling process is taken.

[0061] Compared with existing technologies, the advantages of the drill bit displacement change rate calculation model are that it uses weighted least squares fitting to obtain instantaneous velocity, which has better robustness to measurement noise and abnormal fluctuations; by introducing an acceleration suppression term, it effectively reduces false high-velocity readings caused by instantaneous collision and rebound of the drill bit.

[0062] Furthermore, a threshold for the rate of change of displacement is determined. This threshold is not a fixed value, but is dynamically adjusted based on the performance of the drilling equipment, the estimated strength of the rock mass, and historical drilling data. The model for determining this threshold is as follows:

[0063] ,

[0064] in, The threshold for the rate of change of dynamic displacement; The base threshold is set based on empirical values ​​of typical drilling speeds of drilling rigs in specific rock formations; The impact pressure influence coefficient characterizes the effect of impact energy on the minimum effective drilling speed, and is obtained by fitting experimental data. The root mean square value of the impact pressure in the current monitoring section is calculated from the impact pressure time series data; The rated impact pressure of the drilling rig is used as a reference. The variance of the displacement change rate sequence during the invalid periods is used to estimate the noise level. The noise variance is set based on sensor accuracy and operating conditions.

[0065] Furthermore, the time-series data of the displacement change rate is compared with a dynamic threshold. The magnitude of the value is used to filter out periods where the displacement change rate is consistently higher than the threshold, marking these periods as effective drilling speed periods. Only when a period simultaneously falls within a high-energy window and the drill bit displacement change rate consistently exceeds the threshold is it considered a net drilling process. All time-history monitoring data corresponding to this period are extracted, including impact pressure, rotational pressure, and drill bit position, to form the final net drilling process data.

[0066] In this embodiment, by analyzing the full-process monitoring data of a certain borehole section, the sliding window energy detection method is used to identify high-energy drilling periods. For example, within the time interval [120s, 380s], the window energy value is consistently higher than the upper quartile threshold. Combined with the drill bit displacement change rate, the displacement change rate during this period is greater than the dynamic threshold of 0.15 mm / s, indicating an effective drilling process. After further eliminating invalid intervals such as drill lifting and rod changing, the net drilling process data is extracted, including impact pressure, rotation pressure, and drill bit position time series data, forming a net drilling dataset that can be used for subsequent analysis.

[0067] like Figure 3 The diagram shows the net drilling process of a pneumatic impact drilling rig (curves of drill bit, downward thrust, forward rotation pressure, and impact pressure changing over time). It illustrates a typical net drilling period extracted from the full drilling data, where the impact pressure and drill bit position curves exhibit obvious stage-specific changes, providing crucial data support for subsequent failure mode analysis.

[0068] S3. Using the net drilling process data, extract the first drilling characteristic parameter and the second drilling characteristic parameter.

[0069] In one embodiment, based on the net drilling process data extracted in step S2, key feature parameters for characterizing rock mass failure modes during drilling are calculated through a multi-dimensional feature extraction and quantitative modeling process. The key feature parameter extraction aims to extract stable and sensitive indicators that can distinguish different failure modes from the original monitoring data such as impact pressure, rotation pressure and drill bit position, so as to provide a quantitative basis for subsequent pattern recognition and decision optimization.

[0070] Specifically, firstly, the impact pressure fluctuation intensity is calculated using the impact pressure time-series data from the net drilling process data, serving as the first drilling characteristic parameter. The impact pressure fluctuation intensity is used to quantify the dynamic changes in impact pressure during drilling, comprehensively reflecting the dispersion, range influence, and instantaneous fluctuation amplitude of the pressure value. The calculation model for the impact pressure fluctuation intensity satisfies the following expression:

[0071] ,

[0072] in, The intensity of the impact pressure fluctuation; For the first The impact pressure values ​​at each sampling point were obtained directly from the net drilling process data; To calculate the average impact pressure within the window; The length of the sliding window is set according to the main period of the impact. The range influence factor is used to adjust the contribution of the pressure range to the fluctuation intensity, and is obtained through regression analysis of typical failure mode data. , These represent the maximum and minimum values ​​of the impact pressure within the window, respectively. The pressure normalization coefficient is used to map the range to a reasonable range and is taken as a certain proportion of the rated impact pressure of the drilling rig. The instantaneous change weighting coefficient is used to enhance the contribution of pressure changes between adjacent sampling points to the fluctuation intensity. It is determined through correlation analysis between pressure changes and rock mass fracturing events.

[0073] Compared with existing technologies, the improvement of the impact pressure fluctuation intensity calculation model is that it not only considers the statistical discreteness of pressure, but also introduces the nonlinear effect of the range through the hyperbolic tangent function, and captures the instantaneous fluctuation characteristics through the first-order difference term, so as to more comprehensively and sensitively reflect the dynamic characteristics of impact pressure under different failure modes, such as crushing and cutting.

[0074] Furthermore, using the drill bit position time-series data from the net drilling process data, a displacement mutation intensity calculation model is constructed, and the displacement mutation intensity is obtained through the model as a second drilling characteristic parameter. The displacement mutation intensity is used to quantify sudden, discontinuous displacement events that occur during drilling, which are often closely related to destructive phenomena such as rock mass collapse and crack penetration. The displacement mutation intensity calculation model satisfies the following expression:

[0075] ,

[0076] in, The intensity of the sudden displacement change; for The drill bit position value at any given time is obtained from the net drilling process data; The sampling time interval; To calculate the number of sampling points within the window; This represents the mean of the drill bit position sequence within the window. , These are the mean and variance of the drill bit position sequence throughout the entire net drilling process, respectively, used to globally normalize the position level of the current window to suppress deviations caused by different absolute drilling depths.

[0077] Compared with existing technologies, the advantages of the displacement mutation intensity calculation model are as follows: by calculating the average absolute value of acceleration, it directly reflects the frequency and amplitude of displacement mutation, which can effectively capture sudden sinking or jumping of the drill bit caused by rock collapse, etc. At the same time, the introduction of an exponential decay term based on global statistics can adaptively suppress small fluctuations in the normal uniform drilling range, thereby improving the signal-to-noise ratio for detecting abnormal mutation events.

[0078] Furthermore, regarding the first drilling characteristic parameter, namely the impact pressure fluctuation intensity... and the second drilling characteristic parameter, namely the displacement abrupt change intensity Standardization is performed to eliminate the influence of units and adapt to the requirements of subsequent pattern recognition models on the range of input data; specifically, the Z-score standardization method is used:

[0079] ,

[0080] in, and These are the standardized impact pressure fluctuation intensity and displacement abrupt change intensity, respectively. , The entire diamond cleaning process The mean and standard deviation of the sequence; , These are the mean and standard deviation of the entire net drilling process sequence, respectively; finally, the standardized characteristic parameter sequence is obtained.

[0081] In this embodiment, based on the net drilling process data obtained in step S2, the impact pressure fluctuation intensity was calculated to be 0.48 MPa, and the displacement mutation intensity was calculated to be 0.023 m / s². 2 Subsequently, these two feature parameters were standardized to obtain a standardized impact pressure fluctuation intensity of 1.25 and a displacement mutation intensity of 0.87, which were used as input features for subsequent pattern recognition.

[0082] S4. Based on the first drilling characteristic parameters and the second drilling characteristic parameters, and through the first judgment condition, determine the preliminary damage mode identification result.

[0083] In one embodiment, the first drilling characteristic parameters extracted and standardized in step S3 are used. Second drilling characteristic parameters By constructing a confidence assessment and arbitration mechanism, the initial identification of different failure modes during pneumatic impact drilling is achieved. This initial failure mode identification aims to quickly and accurately classify the drilling process into three typical failure modes: fracturing mode, cutting mode (with obvious pauses), and cutting mode (without obvious pauses), providing an initial basis for subsequent optimization identification and decision-making. The fracturing mode refers to the large-volume fracturing of rock caused by the drill bit's impact, accompanied by severe vibration and impact during the drilling process. The cutting mode (with obvious pauses) refers to the periodic pressing and shearing of the rock by the drill bit teeth, resulting in obvious jumps or pauses in the drilling process. The cutting mode (without obvious pauses) refers to the continuous and stable pressing and shearing of the rock by the drill bit teeth, with a relatively continuous and uniform drilling process.

[0084] Specifically, firstly, based on the first drilling characteristic parameters and the second drilling characteristic parameters, the confidence scores for the breaking mode and the cutting mode are calculated respectively. The confidence score calculation model comprehensively considers the degree of matching between the characteristic parameters and the typical characteristics of each failure mode, and adopts a weighted probability mapping method based on Mahalanobis distance, the expression of which is as follows:

[0085] ,

[0086] ,

[0087] in, , These are the confidence levels for the breakout mode and the entry mode, respectively. ,

[0088] , , In smash mode and The mean and variance were obtained through statistical analysis of a historical typical pattern database; , , , In the cut-in mode and The mean and variance; , These are the characteristic weight coefficients, which reflect the contribution of the impact pressure fluctuation intensity and the displacement mutation intensity to the mode discrimination, respectively, and are determined through characteristic importance analysis.

[0089] Furthermore, based on the confidence levels of the smashing mode and the cutting mode, a confidence difference ratio is calculated to quantify the distinguishability of the two modes; the expression for calculating the confidence difference ratio is:

[0090] ,

[0091] in, The confidence difference ratio is the ratio of confidence levels. The closer the value is to 1, the higher the discrimination between the two modes.

[0092] Further, determine the first preset threshold. This threshold is determined through confusion matrix analysis of a large number of samples and is used to determine whether pattern determination can be made directly based on confidence level.

[0093] Furthermore, compare the confidence level difference ratios. With the first preset threshold Size:

[0094] like If the pattern with high confidence is identified as the preliminary destruction pattern recognition result; that is, if If it is, it is determined to be the smashing mode; otherwise, it is determined to be the cutting mode.

[0095] like If this occurs, the feature-based clustering result arbitration mechanism will be triggered, and the pattern to which the cluster center belongs will be determined as a preliminary violation of the pattern recognition result.

[0096] The result arbitration mechanism based on feature clustering specifically includes: arbitrating the current feature points... Cluster analysis is performed on the feature samples in the historical typical pattern database. The K nearest neighbor algorithm is used to find the K nearest neighbor samples of the current point. The number of smashing and cutting patterns in the K nearest neighbor samples is counted. The pattern with the dominant number is determined as the preliminary destruction pattern recognition result of the current point.

[0097] Furthermore, the preliminary damage pattern identification results for each analysis period are output, including the pattern type and its confidence level, for use in subsequent optimization processes.

[0098] In this embodiment, based on the standardized feature parameters, the confidence level of the shattering pattern is calculated to be 0.82, the confidence level of the cutting-in pattern is 0.23, and the confidence level difference ratio is 0.72, which is higher than the preset threshold of 0.6. Therefore, this time period is directly determined to be the shattering pattern. In another time period, the confidence level of the shattering pattern is 0.45, the confidence level of the cutting-in pattern is 0.43, and the difference ratio is lower than the threshold, triggering the clustering arbitration mechanism. Based on the fact that the cutting-in pattern is dominant among the K-nearest neighbor samples, it is finally determined to be the cutting-in pattern (with obvious pauses).

[0099] S5. Based on the preliminary damage mode identification results, determine the optimized damage mode identification results through the second judgment condition.

[0100] In one embodiment, based on the preliminary damage pattern recognition results obtained in step S4, the preliminary results are refined and optimized by introducing a second judgment condition of uncertainty quantification assessment and multi-source consistency verification, thereby improving the accuracy and robustness of pattern recognition. The optimization of damage pattern recognition aims to solve the problem of unreliable preliminary recognition results caused by data noise, feature overlap or sudden changes in operating conditions, and ensure that the final output damage pattern recognition results have high confidence and engineering applicability.

[0101] Specifically, the location uncertainty quantification value and temporal continuity index of the preliminary damage pattern identification result are first obtained. The calculation model for the location uncertainty quantification value is as follows:

[0102] ,

[0103] in, This is a quantified value for positional uncertainty, reflecting the degree of dispersion of the current feature point distribution in the feature space; This is the number of local samples used to calculate uncertainty; , For the first Feature values ​​of neighboring samples; , The mean value of the feature parameters within the current window; the calculation model for the time series continuity index is as follows:

[0104] ,

[0105] in, It is a temporal continuity indicator, reflecting the consistency of pattern recognition results in adjacent time periods; This represents the total number of time periods analyzed. For the first Preliminary identification pattern for time periods; For indicator functions, when If the value is 1, then the value is 0; otherwise, the value is 0.

[0106] Further, determine the second preset threshold. This threshold is determined by statistical analysis of the uncertainty distribution of historical identification results and is used to determine whether preliminary results can be directly adopted.

[0107] Furthermore, compare the quantified values ​​of the positional uncertainty. With the second preset threshold Size:

[0108] like This indicates that the preliminary identification results have high local consistency, and the preliminary damage mode identification results are directly output as the optimized damage mode identification results.

[0109] like If this is the case, the result optimization process based on the second judgment condition will be triggered, and the following steps will be executed further:

[0110] The consistency score between drilling characteristic parameters and a historical typical pattern database is calculated, and the calculation model for the consistency score is as follows:

[0111] ,

[0112] in, For consistency score; The number of samples in the historical typical pattern database that are the same as the current preliminary pattern; , For the first in the database Standardized features of each sample; This is the covariance matrix of the corresponding samples; and These are the standardized impact pressure fluctuation intensity and displacement abrupt change intensity, respectively.

[0113] Determine the third preset threshold This threshold is determined by the recall-precision curve of historical samples;

[0114] Compare the consistency scores With the third preset threshold Size:

[0115] like This indicates that the current characteristics are highly consistent with historical typical patterns, and the preliminary damage pattern identification results are used as the optimized damage pattern identification results;

[0116] like Then, feature reconstruction and pattern re-identification based on graph neural networks are initiated, and the re-identification result is used to optimize and destroy the pattern identification result. The feature reconstruction and pattern re-identification based on graph neural networks includes the following sub-steps:

[0117] Feature nodes are constructed using the first drilling feature parameter, the second drilling feature parameter, and the average value of the rotary pressure. An undirected weighted graph, i.e. a feature heterogeneous graph, is constructed using the mutual information between features as edge weights.

[0118] A graph attention network is used to aggregate and enhance the node features of the heterogeneous feature graph to obtain a reconstructed high-dimensional feature vector.

[0119] The reconstructed high-dimensional feature vector is input into a pre-trained pattern classifier to obtain the re-identified pattern and its corresponding probability.

[0120] Based on the probability of the re-identified pattern, the final arbitration is carried out in combination with the expert rule base, and the optimized destruction pattern identification result is output. That is, the optimized destruction pattern identification result corresponding to each analysis period is output, including pattern type, confidence level, optimization flag and suggested confidence level.

[0121] In this embodiment, for the time period initially identified as the smashing pattern, its position uncertainty value is calculated to be 0.12, which is lower than the threshold of 0.20, indicating that the result is reliable and is directly adopted as the optimized result; in another time period, the uncertainty value is 0.35, which is higher than the threshold, and the consistency score is further calculated to be 0.58, which is lower than the threshold of 0.70, triggering the graph neural network re-identification process. The final output optimization result is the cutting pattern (without obvious pauses), and it is marked as the optimized result with a high credibility level.

[0122] S6. Based on the optimized failure mode identification results, different failure modes of aerodynamic impact are determined.

[0123] In one embodiment, based on the optimized failure mode recognition results obtained in step S5, combined with real-time drilling characteristic parameters and a preset decision knowledge base, a drilling status report and personalized decision suggestions for engineering applications are generated. This step aims to transform the pattern recognition results into practical information that can guide on-site drilling operations, including parameter optimization, risk assessment, and wall protection suggestions, forming a complete closed loop from perception to decision-making.

[0124] Specifically, firstly, based on the optimized failure mode identification results, a corresponding drilling parameter optimization strategy library is obtained; the strategy library is a lookup table constructed based on historical data, expert experience, and physical models, which stores the recommended baseline thrust under different failure modes. Reference rotation speed And other relevant parameters, Indicates the destruction mode type, such as smash mode, cut-in mode (with obvious pause), and cut-in mode (without obvious pause).

[0125] Furthermore, based on the drilling parameter optimization strategy library and real-time drilling characteristics, the final recommended drilling parameters are calculated through an adaptive weight allocation model. This adaptive weight allocation model comprehensively considers the current failure mode, real-time characteristic fluctuations, and environmental adaptive factors to achieve dynamic and refined adjustment of the drilling parameters. The adaptive weight allocation model satisfies the following expression:

[0126] ,

[0127] ,

[0128] in, and These are the recommended downward thrust and recommended slewing speed, respectively. For the first The dynamic weights of the damage modes are allocated based on the probability distribution of the current optimized damage mode identification results; and For the first The baseline parameters for each failure mode; , , This is an environmental adaptive adjustment coefficient that is updated in real time based on rock mass characteristics and drilling rig status. and These represent the impact pressure fluctuation intensity and displacement abrupt change intensity of the current window, respectively. and The characteristic reference value is taken from the statistical median under typical working conditions;

[0129] Furthermore, a drilling status report is constructed, which is a structured document containing the following content:

[0130] Optimize the damage pattern identification results: including pattern type, confidence level, and optimization flag;

[0131] Recommended drilling parameters: including thrust, rotation speed and other auxiliary parameters;

[0132] Real-time risk assessment levels: Based on damage modes and characteristic parameters, low-risk, medium-risk, and high-risk levels are classified.

[0133] Recommendations for wall protection: Based on the failure mode, such as the crushing mode which is prone to causing borehole collapse, corresponding wall protection measures are recommended, such as increasing mud viscosity and adjusting casing depth.

[0134] Furthermore, the drilling status report and decision suggestions are displayed to the operator in real time through a human-machine interface, and data export and historical record query are supported. At the same time, the recommended parameters are transmitted to the drilling rig control system through the communication interface, supporting manual setting or automatic adjustment, so as to realize intelligent and precise control of the drilling process.

[0135] In this embodiment, based on the optimized identification results, for the breaking mode period, the recommended thrust is 85 kN and the rotation speed is 28 rpm, with a medium risk level indicated, and it is suggested to increase the mud viscosity to prevent borehole collapse; for the cutting mode (without obvious pauses) period, the recommended thrust is 70 kN and the rotation speed is 32 rpm, which is assessed as low risk and no special wall protection measures are required; finally, a drilling status report is generated, including mode type, recommended parameters, risk level and wall protection suggestions, and is displayed in real time through a human-machine interface, supporting drilling rig parameter adjustment and historical record backtracking.

[0136] like Figure 4 As shown, the curves of drill bit position, downward pressure, and impact pressure changing with time during the net drilling process are presented. The intervals A, B, and C are identified as the drilling mode of the pneumatic impact drill breaking the rock of the working face, corresponding to obvious hole collapse. Through grouting tests, it is verified that the net grouting volume of the borehole is significantly higher than the theoretical value, confirming the accuracy of the identification results and providing a direct basis for the selection of wall protection and parameter optimization.

[0137] Ultimately, through the sequential execution of steps S1 to S6, the entire process from raw data acquisition, net drilling process extraction, feature parameter calculation, preliminary pattern identification, result optimization and confirmation to final decision output was automated, thus achieving accurate determination of different damage modes of aerodynamic impact.

[0138] In summary, this invention, based on time-series monitoring data of the drilling process, synchronously collects multi-dimensional time-series data such as impact pressure, rotational pressure, and drill bit position using high-precision sensors. It employs an improved window energy calculation model and dynamic threshold strategy to accurately extract the net drilling process, and then constructs dual feature parameters of impact pressure fluctuation intensity and displacement mutation intensity. Combined with multi-level discrimination mechanisms such as confidence assessment, cluster arbitration, and graph neural network re-identification, it achieves accurate identification and optimized confirmation of three typical failure modes in pneumatic impact drilling: breakage mode, cutting mode (with obvious pauses), and cutting mode (without obvious pauses). Finally, through adaptive parameter recommendation and risk assessment, it generates drilling status reports and decision suggestions, effectively improving the intelligence level and engineering applicability of hole quality evaluation, drilling parameter optimization, prevention of stuck drill bits and drill bit detachment, and wall protection selection.

[0139] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for determining different failure modes of aerodynamic impact based on borehole progress monitoring, characterized in that, The method includes the following steps: Acquire time-series monitoring data of the drilling process, including time-series data of impact pressure, rotation pressure, and drill bit position; Based on the aforementioned time-history monitoring data, the net drilling process is extracted using an improved window energy calculation model to obtain net drilling process data; Using the net drilling process data, extract the first drilling characteristic parameter and the second drilling characteristic parameter; Based on the first drilling characteristic parameters and the second drilling characteristic parameters, the preliminary damage mode identification result is determined through the first judgment condition; Based on the preliminary damage pattern identification results, the optimized damage pattern identification results are determined through the second judgment condition; The optimized failure mode identification results enable the determination of different failure modes of aerodynamic impact. Based on the aforementioned time-history monitoring data, the net drilling process is extracted using an improved window energy calculation model, yielding net drilling process data including: Based on the time-history monitoring data, an improved window energy calculation model is constructed by introducing a weighted power spectral density integral and a transient impact event enhancement mechanism; the window energy value is obtained using the improved window energy calculation model. Based on the window energy value, the high-energy drilling sampling window is determined using the adaptive dual threshold method; Based on the high-energy drilling sampling window, a calculation model for the rate of change of drill bit displacement is constructed by introducing a sliding fitting slope and an acceleration suppression mechanism. The drill bit displacement change rate is calculated using the aforementioned drill bit displacement change rate calculation model. Based on the performance of drilling equipment, the estimated strength of rock mass, and historical drilling data, a displacement change rate threshold determination model is established; the displacement change rate threshold is calculated using the displacement change rate threshold determination model. Compare the drill bit displacement change rate with the displacement change rate threshold. If the drill bit displacement change rate is consistently higher than the displacement change rate threshold, it is determined to be a net drilling process. Based on the aforementioned drilling process, time-series monitoring data within the time period are extracted to obtain drilling process data; The improved window energy calculation model satisfies the following expression: ， in, For a moment The enhanced energy value of the window in question. This represents the number of sampling points within the sliding window. For the first in the window Impact pressure values ​​at each sampling point For frequency domain energy weighting coefficients, For frequency The weight function, For a moment Power spectral density of nearby signals, , These are the lower and upper limits of the characteristic frequency band of the effective drilling signal, respectively. For the Nyquist frequency, For transient impact weighting coefficients, This is a set of indices for transient impact events detected within the window. For the first The amplitude of the impact pressure change in a transient impact event.

2. The method for determining different failure modes of aerodynamic impact based on borehole progress monitoring according to claim 1, characterized in that, Using the net drilling process data, the extraction of the first drilling characteristic parameter and the second drilling characteristic parameter includes: Using the net drilling process data, an impact pressure fluctuation intensity calculation model is established; using the impact pressure fluctuation intensity calculation model, a first drilling characteristic parameter is calculated, the first drilling characteristic parameter including the impact pressure fluctuation intensity; Using the net drilling process data, a displacement mutation intensity calculation model is constructed; using the displacement mutation intensity calculation model, a second drilling characteristic parameter is calculated, the second drilling characteristic parameter including the displacement mutation intensity; The first drilling characteristic parameter and the second drilling characteristic parameter are standardized to obtain the standardized first drilling characteristic parameter and the standardized second drilling characteristic parameter.

3. The method for determining different failure modes of aerodynamic impact based on borehole progress monitoring according to claim 2, characterized in that, The calculation model for the intensity of impact pressure fluctuations satisfies the following relationship: ， in, The intensity of the impact pressure fluctuation. For the first Impact pressure values ​​at each sampling point To calculate the mean impact pressure within the window, The length of the sliding window. The range impact factor, , These represent the maximum and minimum values ​​of the impact pressure within the window, respectively. This is the pressure normalization coefficient. The instantaneous change weighting coefficients, and the displacement mutation intensity calculation model satisfy the following expression: ， in, The intensity of the sudden displacement. for The drill bit position value at any given time. The sampling time interval, To calculate the number of sampling points within the window, This represents the mean of the drill bit position sequence within the window. , These are the mean and variance of the drill bit position sequence throughout the entire net drilling process, respectively.

4. The method for determining different failure modes of aerodynamic impact based on borehole progress monitoring according to claim 1, characterized in that, Based on the first drilling characteristic parameters and the second drilling characteristic parameters, and through the first judgment condition, the preliminary failure mode identification result is determined to include: Based on the first drilling characteristic parameter and the second drilling characteristic parameter, a calculation model for the confidence level of the breaking mode and the confidence level of the cutting mode is established; the confidence level of the breaking mode and the confidence level of the cutting mode are calculated using the calculation model of the confidence level of the breaking mode and the confidence level of the cutting mode. Based on the confidence levels of the breaking mode and the cutting mode, the confidence level difference ratio is calculated; Determine the first preset threshold; The confidence difference ratio is compared with the first preset threshold. If the confidence difference ratio is greater than the first preset threshold, the pattern with high confidence is determined as the preliminary destruction pattern identification result. If the confidence difference ratio is not greater than the first preset threshold, the result arbitration mechanism based on feature clustering is triggered, and the pattern to which the cluster center belongs is determined as the preliminary destruction pattern identification result.

5. The method for determining different failure modes of aerodynamic impact based on borehole progress monitoring according to claim 1, characterized in that, Based on the preliminary damage pattern identification results, the optimized damage pattern identification results are determined through the second judgment condition, including: Based on the preliminary damage pattern identification results, a location uncertainty quantification value calculation model is established; the location uncertainty quantification value is determined through the location uncertainty quantification value calculation model. Based on the preliminary damage pattern identification results, a time-series continuity index calculation model is established; the time-series continuity index is determined through the time-series continuity index calculation model. Determine the second preset threshold; The position uncertainty quantization value is compared with the second preset threshold. If the position uncertainty quantization value is not greater than the second preset threshold, the preliminary damage pattern recognition result is used as the optimized damage pattern recognition result. If the position uncertainty quantization value is greater than the second preset threshold, the result optimization process based on the second judgment condition is triggered. The optimized destruction mode identification result is determined based on the comparison result between the quantified value of the location uncertainty and the second preset threshold.

6. The method for determining different failure modes of aerodynamic impact based on borehole progress monitoring according to claim 5, characterized in that, The result optimization process based on the second judgment condition includes: Calculate the consistency score between drilling characteristic parameters and the historical typical pattern database; Determine the third preset threshold; Compare the consistency score with the third preset threshold: if the consistency score is not less than the third preset threshold, the preliminary destruction pattern recognition result is used as the optimized destruction pattern recognition result; if the consistency score is less than the third preset threshold, feature reconstruction and pattern re-recognition based on graph neural network are initiated, and the re-recognition result is used as the optimized destruction pattern recognition result.

7. The method for determining different failure modes of aerodynamic impact based on borehole progress monitoring according to claim 6, characterized in that, The feature reconstruction and pattern re-identification based on graph neural networks includes: Construct a feature heterogeneity graph; A graph attention network is used to aggregate and enhance the node features of the heterogeneous feature graph to obtain a reconstructed high-dimensional feature vector. The high-dimensional feature vector is input into a pre-trained pattern classifier to obtain the probability of re-identifying the pattern; Based on the probability of the re-identified pattern, a final arbitration is conducted using an expert rule base to output an optimized destruction pattern identification result. The optimized destruction pattern identification result includes pattern type, confidence level, optimization flag, and suggested confidence level.

8. The method for determining different failure modes of aerodynamic impact based on borehole progress monitoring according to claim 1, characterized in that, The determination of different failure modes of aerodynamic impact is achieved through the optimized failure mode identification results, including: Based on the optimized failure mode identification results, a drilling parameter optimization strategy library is obtained; An adaptive weight allocation model is established based on the drilling parameter optimization strategy library. Using the adaptive weight allocation model, recommended drilling parameters are calculated, including recommended thrust and recommended rotation speed. Based on the recommended drilling parameters, a drilling status report is constructed, which includes optimized failure mode identification results, recommended drilling parameters, real-time risk assessment level, and wall protection recommendations. The drilling status report enables the determination of different damage modes caused by aerodynamic impact.

Images