Parameter acquisition method, medium and device of a hole drilling pole machine

Abstract

The present application relates to the technical field of drilling and erecting pole equipment, and discloses a parameter acquisition method, medium and equipment for a drilling and erecting pole machine. The method acquires multi-source sensing data of a target operation area, which contains geological structure tomography scanning information and real-time operation state parameters of a mechanical power system; performs working condition matching retrieval based on a historical drilling case library, determines a plurality of heterogeneous working condition cluster sets, extracts a plurality of drilling feature peak values and diffusion ranges; takes the feature diffusion range as an analysis scale of a feature extraction network layer, performs time sequence feature slicing processing on the multi-source sensing data, and generates a plurality of drilling state feature sequence sets; performs cross-scale feature interaction fusion on the feature sequence sets, and generates a target fusion drilling state feature matrix; and generates an optimal drilling parameter configuration scheme according to the matrix and a preset drilling parameter constraint condition. The method comprehensively integrates multi-source information, accurately extracts working condition features, realizes optimal configuration of parameters, and is suitable for parameter acquisition of drilling and erecting pole equipment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of drilling and erecting pole equipment, in particular to a parameter acquisition method, medium and device of a drilling and erecting pole machine. BACKGROUND

[0002] In the infrastructure construction of power, communication, construction and other industries, drilling and erecting pole operation is a key process, and its operation efficiency and construction quality directly affect the progress of subsequent projects. As the core equipment for completing the operation, the rationality of the parameter configuration of the drilling and erecting pole machine plays a crucial role in drilling accuracy, pole stability and equipment service life.

[0003] The traditional parameter acquisition method of the drilling and erecting pole machine mostly relies on the experience judgment of the operator, and manually sets the drilling speed, feed amount, torque and other parameters according to the approximate geological conditions of the operation area and the equipment model. This method has obvious limitations. Due to the large difference in geological structure of different operation areas, even in the same area, there may be uneven stratum distribution. It is difficult to accurately grasp the actual working condition in the drilling process only by experience, which may lead to unreasonable parameter configuration, and further cause low drilling efficiency, equipment wear and tear, and even safety hazards such as hole wall collapse and pole inclination.

[0004] With the development of sensing technology, some drilling and erecting pole equipment begins to be equipped with a single type of sensor for collecting part of the parameters in the drilling process, such as drilling speed, torque, etc., and making simple parameter adjustment based on these data. However, the working condition information reflected by single sensing data is limited and cannot fully cover factors such as geological structure and mechanical operating state. For example, it is difficult to accurately judge the hardness change of the stratum by only using torque data, and it is also difficult to evaluate the stability of the hole wall by only relying on the drilling speed. In addition, the processing method of sensing data in the prior art is relatively simple, mostly using direct comparison or linear fitting method, lacking deep mining and multi-dimensional analysis of data, and it is difficult to extract characteristic information that can accurately reflect complex working conditions, resulting in the timeliness and accuracy of parameter adjustment still to be improved.

[0005] In actual operation, the drilling process is a dynamic process, and the geological conditions and mechanical operating state will change with the increase of drilling depth, requiring real-time acquisition and analysis of multi-source information for dynamic adjustment of drilling parameters. However, the traditional parameter acquisition method and the existing part of the improved scheme mostly fail to establish an effective multi-source data fusion mechanism and dynamic analysis model, and cannot realize real-time and accurate control of the drilling process, making it difficult to meet the needs of efficient and safe construction under complex geological conditions. Therefore, developing a method that can comprehensively utilize multi-source sensing data, accurately extract working condition characteristics, and realize optimal parameter configuration has become an important direction for the development of current drilling and erecting pole machine technology. SUMMARY

[0006] The present application aims to provide a parameter acquisition method, medium and equipment of a drilling pole machine to solve the problems raised in the background.

[0007] To achieve the above-mentioned purpose, the present application provides a parameter acquisition method of a drilling pole machine, which comprises:

[0008] Obtain multi-source sensing data of a target working area, which contains geological structure tomography information and real-time running state parameters of a mechanical power system;

[0009] Perform working condition matching retrieval based on a historical drilling case library to determine a plurality of heterogeneous working condition cluster sets, traverse the plurality of heterogeneous working condition cluster sets to perform feature set extraction, and obtain a plurality of drilling feature peak values and a plurality of drilling feature diffusion ranges;

[0010] Use the plurality of drilling feature diffusion ranges as analysis scales of a plurality of feature extraction network layers, perform time series feature slicing processing on the multi-source sensing data through the plurality of feature extraction network layers, and generate a plurality of drilling state feature sequence sets;

[0011] Perform cross-scale feature interaction fusion on the plurality of drilling state feature sequence sets to generate a target fusion drilling state feature matrix;

[0012] Generate an optimal drilling parameter configuration scheme according to the target fusion drilling state feature matrix and a preset drilling parameter constraint condition.

[0013] Preferably, the working condition matching retrieval based on the historical drilling case library to determine the plurality of heterogeneous working condition cluster sets comprises:

[0014] Extract a geological structure feature vector and a mechanical power system state feature vector in the multi-source sensing data;

[0015] Calculate the Euclidean distance between the geological structure feature vector and the case geological feature in the historical drilling case library, and screen a candidate case subset that meets a similarity threshold;

[0016] Perform energy spectrum density distribution analysis on the mechanical power system state feature vector in the candidate case subset to generate a working condition clustering center point set;

[0017] Divide the heterogeneous working conditions of the historical drilling case library according to the working condition clustering center point set to generate a plurality of heterogeneous working condition cluster sets containing geological attribute labels and power state labels.

[0018] Preferably, the traversing of the plurality of heterogeneous working condition cluster sets to perform feature set extraction to obtain a plurality of drilling feature peak values and a plurality of drilling feature diffusion ranges comprises:

[0019] extract a drilling resistance instantaneous change amount set from the multiple heterogeneous working condition cluster sets, and perform gradient mutation point detection on the drilling resistance instantaneous change amount set;

[0020] identify a space-time coordinate corresponding to the gradient mutation point, and generate a drilling feature peak space-time distribution map;

[0021] using the drilling feature peak space-time distribution map as an index, performing feature correlation diffusion analysis on the multiple heterogeneous working condition cluster sets, calculating an influence radius mean value of each gradient mutation point, and generating the multiple drilling feature diffusion ranges.

[0022] Preferably, the time series feature slicing processing of the multiple-source sensing data by the multiple feature extraction network layers to generate the multiple drilling state feature sequence sets comprises:

[0023] According to the multiple drilling feature diffusion ranges, a sliding window size parameter is set to perform time dimension segmentation sampling on the multiple-source sensing data;

[0024] The segmented sampling data is input into a parallel convolution branch network to extract rock stratum impedance phase gradient features under different time scales;

[0025] The rock stratum impedance phase gradient features are time series aligned and spliced to generate the multiple drilling state feature sequence sets containing vibration spectrum features and drill bit wear state features.

[0026] Preferably, the cross-scale feature interaction fusion of the multiple drilling state feature sequence sets to generate a target fusion drilling state feature matrix comprises:

[0027] The first drilling state feature sequence set and the second drilling state feature sequence set are selected to perform feature similarity measurement to generate a feature correlation weight coefficient matrix;

[0028] According to the feature correlation weight coefficient matrix, the second drilling state feature sequence set is weighted and convoluted to generate a primary fusion feature tensor;

[0029] The primary fusion feature tensor and the third drilling state feature sequence set are cross-scale feature interacted, and the feature interaction operation is iteratively executed until the fusion of all drilling state feature sequence sets is completed, and the target fusion drilling state feature matrix is output.

[0030] Preferably, the generation of the optimal drilling parameter configuration scheme according to the target fusion drilling state feature matrix and a preset drilling parameter constraint condition comprises:

[0031] Analyzing the semantic dependency relationship in the drilling task instruction, extracting a drilling depth priority coefficient and a verticality error tolerance threshold;

[0032] mapping the drilling depth priority coefficient as a parameter optimization weight factor, and combining the target fusion drilling state feature matrix to construct a multi-objective optimization function;

[0033] A non-dominated sorting strategy is used to perform a Pareto frontier search on the rotational speed parameter range, the feed pressure threshold value and the cooling liquid flow interval to generate a candidate parameter configuration set;

[0034] According to the perpendicularity error tolerance threshold, the candidate parameter configuration set is filtered for feasibility, and the optimal drilling parameter configuration scheme is output.

[0035] Preferably, the method further comprises:

[0036] Real-time monitoring of the dynamic deviation during drilling is performed, and when the deviation of the drill rod attitude angle exceeds a preset threshold, a parameter reconfiguration process is triggered;

[0037] According to the mapping relationship between the current drilling depth coordinate and the geological structure tomography information, the rock hardness compensation coefficient in the drilling parameter constraint condition is updated;

[0038] Based on the updated rock hardness compensation coefficient, the multi-objective optimization function is recalculated to generate a dynamically adjusted optimal drilling parameter configuration scheme.

[0039] Preferably, the parameter reconfiguration process comprises:

[0040] Collecting drill rod vibration modal spectral data, and identifying abnormal resonance frequency bands from the drill rod vibration modal spectral data;

[0041] According to the matching of the center frequency of the abnormal resonance frequency band with the mechanical resonance characteristics in historical fault cases, a risk level evaluation index is generated;

[0042] When the risk level evaluation index exceeds a safety threshold, a rotational speed frequency reduction constraint and a feed force attenuation coefficient are injected into the optimal drilling parameter configuration scheme.

[0043] Preferably, the present application further comprises a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the above method.

[0044] Preferably, the present application further comprises a drilling rod machine parameter acquisition device, which comprises a processor and a memory, the memory stores a computer program, and the processor executes the computer program to implement the steps of the above method.

[0045] Compared with the prior art, the present application has the following advantages:

[0046] By acquiring multi-source sensing data of the target operation area, covering geological structure tomography information and real-time operation state parameters of the mechanical power system, various information in the operation process can be comprehensively captured from multiple dimensions, breaking the limitations of traditional single data source, and making the cognition of working conditions more comprehensive and in-depth. Geological structure information can reflect the distribution, hardness, integrity and other key characteristics of the stratum, while the mechanical operation state parameters can reflect the load, speed, vibration and other conditions of the equipment. The combination of the two provides a rich basis for subsequent parameter configuration.

[0047] Based on the historical drilling case library, the working condition matching retrieval is performed to determine multiple heterogeneous working condition cluster sets and extract feature peaks and diffusion ranges, fully utilizing the value of historical experience data. The historical cases contain operation rules and characteristics under different geological conditions and different equipment states. Through clustering analysis and feature extraction, complex working condition information can be converted into representative feature indexes, which can accurately reflect the key features and change ranges under different working conditions, providing clear direction and scale reference for subsequent feature analysis, making the judgment of the current operation condition more targeted and accurate, and avoiding the blindness of judgment based on experience or single data.

[0048] Multiple drilling feature diffusion ranges are used as the analysis scale of multiple feature extraction network layers to perform time series feature slicing processing on multi-source sensing data, generating multiple drilling state feature sequence sets, and realizing fine processing of data. Different feature diffusion ranges correspond to different analysis granularities. Through multi-layer network structure for time series slicing of data, dynamic change features under different time scales in the drilling process can be captured, whether it is short-term instantaneous fluctuation or long-term trend change. This multi-scale feature extraction method makes the dynamic change perception of working conditions more sensitive, and can timely discover the subtle changes in the drilling process, providing detailed feature support for parameter adjustment.

[0049] Cross-scale feature interaction fusion is performed on the multiple drilling state feature sequence sets to generate a target fusion drilling state feature matrix, further improving the quality of feature information. Cross-scale fusion can organically combine feature information of different levels and dimensions, eliminate redundancy and conflicts between features, and strengthen useful feature signals, so that the fused feature matrix can more accurately and comprehensively reflect the overall working condition in the drilling process. This fusion not only retains the local information of each single feature, but also excavates the correlation between different features, providing more valuable feature input for subsequent parameter generation.

[0050] According to the target fusion drilling state feature matrix and the preset drilling parameter constraint condition, an optimal drilling parameter configuration scheme is generated, so that the parameter configuration is more suitable for actual operation requirements. The fused feature matrix accurately reflects the current geological conditions and equipment state, and in combination with the preset constraint conditions, such as equipment performance limitation, construction quality standard, etc., an optimal parameter scheme suitable for the current working condition can be generated under the premise of meeting various requirements. This method can adapt to changes in different geological structures and operating environments, ensure that the equipment operation in the drilling process matches the geological conditions, reduce various problems caused by improper parameters, improve the smoothness and stability of the operation, and also help to reduce the invalid loss of the equipment, prolong the service life of the equipment, and ensure the orderly progress of the construction process. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 A working principle diagram of the parameter acquisition method of the drilling pole machine is provided.

[0052] Figure 2 A flowchart for generating a heterogeneous working condition cluster set is provided.

[0053] Figure 3 A flowchart for extracting drilling feature peak value and diffusion range is provided.

[0054] Figure 4 A flowchart for cross-scale feature interaction fusion is provided.

[0055] Figure 5 A flowchart for generating an optimal drilling parameter configuration is provided. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0057] Please refer to Figure 1 The present application provides a parameter acquisition method for a drilling pole machine, which comprises:

[0058] The geological structure tomography information of the target operation area and the real-time running state parameters of the mechanical power system are acquired as multi-source sensing data. Based on the historical drilling case library, a working condition matching retrieval is performed to divide a plurality of heterogeneous working condition cluster sets. The drilling feature peaks and their diffusion ranges are extracted by traversing these sets as the analysis scale of the feature extraction network layer. The multi-source sensing data is processed by a plurality of feature extraction network layers to generate a drilling state feature sequence set. A cross-scale feature interaction fusion operation is performed to generate a target fusion drilling state feature matrix. According to the matrix and the preset drilling parameter constraint condition, an optimal drilling parameter configuration scheme is generated, including the parameter combination of the rotation speed, the feed pressure and the cooling liquid flow.

[0059] Embodiment 1: refer to Figure 2 The geological structure tomography information of the target operation area is acquired by array-type geological radar, the scanning depth resolution is set to 0.1 meters, and three-dimensional point cloud data containing rock layer density distribution is formed. The rock layer density value is converted to kilograms per cubic meter, and the data dimension is 200x200x150. The real-time running state parameters of the mechanical power system are acquired through the CAN bus protocol, the sampling frequency is 500 Hz, and the parameters include the hydraulic motor output torque (unit: Newton-meter), the main shaft radial vibration frequency (0-200 Hz frequency band) and the drill pipe axial deflection angle (unit: degree). The geological structure feature vector is composed of the following dimensions: the rock layer density average value, the density gradient maximum value and the density mutation point number in the ±0.5 meter range of the drilling axis in the density point cloud data are selected; the fracture zone position is defined as the coordinate of the adjacent rock layer interface in the tomographic image where the density difference exceeds 0.5 grams per cubic meter; the underground water permeability is calculated by the resistivity inversion model. The mechanical power system state feature vector extracts the average features from 200 continuous sampling points: the hydraulic motor output torque takes the peak value coefficient after variance normalization, the main shaft vibration frequency takes the amplitude ratio of the first three order resonance frequencies, and the drill pipe deflection angle takes the square root of the standard deviation.

[0060] The historical drilling case library contains 500 completed drilling cases, and the case geological feature vector is stored in the graph database. When calculating the Euclidean distance between the current geological structure feature vector and the case library vector, min-max normalization is performed on each dimension. The similarity threshold is set to 0.85, and the candidate case subset with a distance value below 0.15 (1-0.85) is selected. The size of the subset is controlled within the range of 20%-30% of the total number of cases in the library. The spectral density distribution analysis object is the mechanical power system state feature vector of the candidate case subset: after the hydraulic pressure pulsation signal is processed by the Hanning window function, 1024-point FFT transformation is performed to generate the energy spectrum in the 0-100Hz frequency band; the rotational speed fluctuation spectrum uses the Welch method to calculate the power spectral density, and the frequency band resolution is set to 1Hz. The power spectral density matrix is reduced to a three-dimensional space by principal component analysis, and the K-means clustering algorithm is used to iteratively calculate the centroid position. The number of cluster center points is determined according to the contour coefficient, and the typical value is 3-5. The cluster center point coordinates correspond to specific working conditions: for example, point (0.45, 0.82, 0.13) represents a high-frequency vibration working condition (120-150Hz frequency band energy ratio 60%), and point (0.12, 0.37, 0.89) represents a low-speed high-torque working condition (torque average >1800 Newton·m).

[0061] According to the working condition cluster center point set, the historical drilling case library is divided into different types of working conditions: taking each cluster center point as the core, 150-200 cases with the nearest Euclidean distance in the feature space are labeled. Each case has a dual labeling system: the geological attribute label is classified according to the rock type coding system (such as TAG-7 represents a gravel-containing sandstone layer), and the dynamic state label is generated according to the cluster alias (Cluster-A represents a high-frequency vibration dominant working condition). The different types of working condition cluster sets formed by the division are stored as a graph structure, and the node attributes include drilling depth, drill bit type, drilling time, etc. metadata. Boundary verification is performed between different types of working condition cluster sets: when a case falls into the overlapping area of two cluster sets at the same time, it is re-assigned according to the main state parameter ratio. The final multiple different types of working condition cluster sets have the characteristics of asymmetric distribution: the sandstone cluster set accounts for 40% of the total, the clay layer cluster set accounts for 35%, and the rest are special rock layer combination working conditions. In the dynamic state label distribution, the medium-speed stable working condition label accounts for 55%, and the abnormal working condition label is scattered in the remaining part. The geological attribute label and the dynamic state label establish a cross-index relationship: for example, when the "basalt layer" and "high-frequency vibration" labels are activated at the same time, they are associated with the drill bit alloy abnormal wear warning module.

[0062] Example 2: see Figure 3The instantaneous change set of drilling resistance is extracted from the heterogeneous working condition cluster set, which is derived from the real-time acquisition record of the drill pipe axial pressure sensor. The sensor acquires the resistance value in the direction of the drill bit advancement at a sampling frequency of 100 Hz, and the first derivative is calculated after the original data are processed by the sliding average filter. The instantaneous change set is stored as a three-dimensional array structure in time sequence: the time stamp accuracy is up to milliseconds, the drilling depth coordinate accuracy is up to centimeters, and the resistance change quantity unit is Newton / second. For each working condition cluster set, 50 to 80 groups of case data contained therein are independently processed.

[0063] The gradient mutation point detection adopts an improved Canny operator algorithm. The resistance change rate threshold interval is set to [15, 40] Newton / second, and the detection window length is 10 seconds of data segment. The algorithm execution includes three core operations: calculating the Gaussian smoothed derivative of the resistance change sequence to generate the gradient amplitude distribution graph; applying a double-threshold hysteresis threshold processing, in which the high threshold is set to the 70% quantile value of the gradient amplitude distribution, and the low threshold is the 40% quantile value; performing edge tracking on the mutation points that cross the threshold to connect the discontinuity points to form a complete mutation point set. During the detection process, the parameters are automatically adjusted for different working condition cluster types: the sandstone formation cluster set enables a higher threshold (30 Newton / second), and the clay formation cluster set adopts a lower threshold (20 Newton / second). The successfully identified mutation points are labeled with three attributes: time coordinate (accurate to data point index position), depth coordinate (corresponding to drilling vertical positioning value), and gradient intensity (actual resistance change rate Newton value).

[0064] When generating the spatiotemporal distribution graph of the drilling feature peak value, a three-dimensional coordinate system is used to construct a point cloud model. The horizontal axis represents the time dimension, which is divided into scales in minutes; the vertical axis represents the drilling depth, which is in meters; and the third axis represents the gradient intensity value, which is mapped using a logarithmic scale. Each mutation point in the point cloud model contains a topology relationship matrix, which identifies the adjacent three regular data points before and after it. The spatial distribution model performs Delaunay triangulation to form a spatial correlation network of the mutation points. The distribution graph is labeled with a geological attribute label partition: when the geological attribute label in the heterogeneous working condition cluster set is “basalt layer”, a red highlight identification is superimposed on the corresponding point cloud area.

[0065] The execution feature correlation diffusion analysis adopts an exponential decay model. For each mutation point, the original resistance data within a 2-second time span around it is extracted. The resistance fluctuation variance values of 5 evenly distributed sampling points are calculated, and the sliding window method is used for variance calculation with a window size of 0.5 seconds. The rock hardness parameter comes from the geological structure tomography information database, and the hardness value (Brinell hardness HB unit) is obtained according to the drilling depth coordinate index. A fitting function is established: diffusion influence distance = k x (fluctuation variance / rock hardness)^0.5, where the coefficient k is preset according to different working condition clusters: 0.85 for high-frequency vibration working condition, and 1.2 for conventional working condition. The influence radius is calculated synchronously in time and depth dimensions: the time influence radius is converted according to the number of data points, and the depth influence radius is converted according to the drilling tool per revolution footage.

[0066] The influence radius calculation includes an error correction link. Randomly select 15% of the sample points in the heterogeneous working condition cluster set as the validation set, and measure the actual influence range of the validation points. The measurement method is: expand outward from the mutation point as the center, and record the distance point at which the drilling rod axial vibration energy decays to 80% of the normal value. Calculate the residual sum of squares of the theoretical model output value and the actual distance, and when the residual mean exceeds the preset tolerance (0.01 seconds in time dimension, 0.005 meters in depth dimension), activate parameter adaptive adjustment. Update the exponential term of the exponential decay model dynamically during the correction process: initially set the exponential value to 0.5, and adjust the step size by 0.01 in each iteration.

[0067] The final quantitative form of the diffusion range of multiple drilling features is a numerical interval. A single diffusion range data contains four dimensions: time lower bound (e.g. -1.8 seconds), time upper bound (+1.8 seconds), depth lower bound (-0.42 meters), and depth upper bound (+0.42 meters). The interval boundary values are rounded to two decimal places, and the positive and negative values represent the diffusion range before and after the mutation point coordinate position. The interval effectiveness is confirmed by the historical case backtracking mechanism: search for similar gradient mutation points in the historical drilling case library, and the probability of subsequent engineering problems occurring within the labeled time and space range exceeds 85%. Each heterogeneous working condition cluster set corresponds to an independent diffusion range set, which contains 20 to 35 interval data, sorted in descending order of gradient intensity value and stored in a ring buffer memory. The diffusion range data output format is JSON structure, which includes the metadata header of rock type code and power state label.

[0068] Example 3: see Figure 4The sliding window size parameters are dynamically adjusted based on the borehole feature diffusion range. For depth diffusion ranges within ±0.5 meters, the time window length is set to 3 seconds, and the window sliding step is 0.5 seconds; for time diffusion ranges within ±2 seconds, the window overlap rate is configured to 30%. Multi-source sensor data is segmented using a non-uniform interval strategy, increasing the sampling density to three times the normal value in areas near gradient abrupt change points. The sampled data block standardization process includes two stages: first, Z-score normalization is performed on the data from each sensor channel, and then the range method is used to map the channel values ​​to the [0,1] interval. The segmented data blocks are stored as a four-dimensional tensor structure, with dimensions of time step, number of sensor channels, number of sampling points, and feature dimension, respectively.

[0069] The parallel convolutional branch network architecture comprises two independent processing paths. The first branch network is configured with five one-dimensional convolutional layers, with the kernel size set to five sampling points along the time axis, and the channel expansion sequence being [16, 32, 64, 32, 16]. Each convolutional layer is followed by a LeakyReLU activation function with a fixed slope parameter of 0.01. The second branch network employs a dilated convolutional structure with three convolutional modules, the dilation rate increasing in a sequence of [1, 2, 4], and the base kernel size remaining constant at 3. The output feature maps of both branches are channel-compressed using depthwise separable convolution, reducing the feature dimension from 64 to 32. Phase-locked loop technology is introduced in the extraction process of rock stratum impedance phase gradient features: the vibration sensor signal and the reference signal are orthogonally demodulated, and the instantaneous phase difference is calculated. Phase gradient characteristics Calculated using the following formula:

[0070]

[0071] in: This indicates the fundamental frequency of the drill bit's vibration, measured in Hertz (Hz). express The instantaneous phase difference at any given moment, in radians; The output is a normalized phase gradient change rate, dimensionless. During feature extraction, the 10Hz timescale features mainly reflect changes in the macroscopic structure of the rock strata, while the 50Hz timescale features correspond to the development of microscopic fractures. The feature maps output by the two branch networks undergo temporal alignment: a cubic spline interpolation algorithm is used to downsample the temporal resolution of the 50Hz feature map to 10Hz, and the interpolation node spacing is set to 1 / 5 of the original data points.

[0072] The time-alignment concatenation operation is performed synchronously in three dimensions of the feature space. The vibration spectrum features are extracted from the acceleration sensor signals, generating a time-frequency matrix through short-time Fourier transform with a frequency resolution of 2 Hz and a time window length of 256 points. The drill wear state features are obtained in combination with the temperature sensor data and the acoustic emission signals: the temperature change rate features are calculated in terms of the change in degrees Celsius per minute, and the energy accumulation features of the acoustic emission signals are integrated at intervals of every 10 seconds. The concatenated feature sequences form a three-dimensional structure: the time step dimension remains the same as the original data with 300 points, and the feature channel number is expanded to 128 dimensions, including 32-dimensional phase gradient features, 64-dimensional vibration spectrum features, and 32-dimensional wear state features.

[0073] The feature similarity measure adopts an improved dynamic time warping algorithm. When calculating the similarity matrix of the first set of borehole state features (vibration spectrum features) and the second set of drill wear state features, the slope of the path search is constrained to the interval [0.5, 2]. The similarity weight coefficient The local feature matching degree and the global trend consistency are determined by the product:

[0074]

[0075] Where: represents the feature vector of the first set at time i, represents the feature vector of the second set at time j, and K represents the local trend analysis window half-width, which is set to 5 time points. The generated 12x12 feature correlation weight coefficient matrix is processed by symmetrization, and the diagonal elements are strengthened to 1.2 times the neighborhood mean.

[0076] The weighted convolution fusion process adopts a separable convolution kernel to achieve parameter efficiency optimization. After applying the weight matrix to the second set of borehole state feature sequences, a 3x3 deep convolution kernel is used to process each feature channel independently, and then the channel information is integrated through a 1x1 point-by-point convolution. The step size of the convolution operation is set to 2, and the edge padding mode is selected as mirror reflection. The generation of the primary fusion feature tensor undergoes three feature transformations: first, the feature dimension is expanded to 256 dimensions through a fully connected layer, then the distribution is adjusted through a batch normalization layer, and finally a Swish activation function is applied to complete the nonlinear mapping. The output tensor size is 64xNxN, where N represents the number of time steps, and each spatial position contains a 256-dimensional feature vector.

[0077] The cross-scale feature interaction adopts a gated attention mechanism. The primary fusion feature tensor and the third borehole state feature sequence set (rock fracture distribution features) are input into the interaction module, which includes three processing units: the scale alignment unit unifies the feature map size through bilinear interpolation, the attention generation unit calculates the attention weights in the spatial and channel dimensions, and the feature reorganization unit performs weighted summation. The interaction process is iterated three times: the first interaction focuses on low-frequency feature fusion, and the attention weight is biased towards the spatial low-frequency component; the second interaction enhances high-frequency details, and the attention window is reduced to 1 / 4 of the original size; the third interaction balances global and local features, and adopts a hybrid attention strategy. The final output target fusion borehole state feature matrix contains 512 feature channels, the time resolution remains 1 / 4 of the original input data, and the spatial resolution is reduced to 1 / 8 of the original size through maximum pooling. The matrix storage adopts a sparse format, and the non-zero element threshold is set to 0.05 times the maximum value.

[0078] Example 4: see Figure 5 The borehole task instruction parsing adopts a semantic dependency tree analysis model. The input instruction "drill a 15-meter deep communication rod hole, prioritize construction progress, and the vertical deviation angle should not exceed ±1.5 degrees" is decomposed into a dependency relation graph. The node "depth 15 meters" is attached with a depth priority attribute, and the weight coefficient is converted to a value of 0.93 through a mapping table; the node "vertical deviation angle ±1.5 degrees" extracts the tolerance threshold attribute, and the quantized value is 1.5 degrees. At the same time, the implicit constraint condition is identified: the vibration amplitude limit value is obtained from the standard value 50μm in the industry specification library.

[0079] A target conflict coefficient matrix is introduced when constructing the multi-objective optimization function. Based on the 256-dimensional vector of the target fusion borehole state feature matrix, the drilling efficiency objective function is defined as the product of the drill rod axial footage speed and the current rotation speed, and the drill bit wear objective function adopts the weighted sum of the rate of change of the cutting edge temperature and the stress concentration coefficient. The parameter optimization weight factor is related to the depth priority coefficient: when the coefficient ≥0.9, the drilling efficiency target weight is set to 0.75; when the coefficient <0.9, a linear mapping is used. The parameter constraint boundary value is dynamically adjusted: in the basalt stratum working condition, the upper limit of the rotation speed is reduced to 450rpm; when encountering underground water flow, the lower limit of the cooling fluid flow is increased to 15L / min. The Pareto front search execution process includes population initialization and elite reservation mechanisms. The initial parameter population size is 100 groups, and the generation strategy of the initial parameter population in the multi-objective optimization process includes the sampling method and numerical distribution characteristics of the key operating parameters of the borehole stand.

[0080] Table 1: Parameter initialization configuration table.

[0081] Parameter type Generation method Numerical distribution Rotational speed (rpm) Latin hypercube sampling 100-500 uniform distribution Feed pressure (MPa) Gaussian perturbation Mean 80, standard deviation 15 Coolant flow rate (L / min) Adaptive interval division 5-30 non-uniform segmentation

[0082] The non-dominated sorting process sets two optimization objectives: Objective 1 (drilling efficiency) requires maximization, and Objective 2 (bit wear) requires minimization. Genetic operations with a crossover probability of 0.85 and a mutation probability of 0.15 are performed for each generation iteration. After 50 generations of evolution, 20 sets of non-dominated solutions are screened to form a candidate parameter configuration set. These solution sets form a continuous distribution: high-efficiency solutions are concentrated in the high-rotation-speed zone (400-500 rpm), and low-wear solutions are distributed in the low-pressure zone (< 50 MPa).

[0083] The verticality error simulation calculation uses a multi-physics coupling model. The candidate parameter configuration is input, and the theoretical deviation angle value is calculated through the drill pipe flexibility matrix and the formation reaction force tensor. The drill pipe is modeled as a hollow cylinder with a diameter of 89 mm and a wall thickness of 8 mm, and the material elastic modulus is 210 GPa. The geological reaction force is obtained by interpolation according to the rock compressive modulus, with an interpolation point interval of 0.1 meters. The deviation angle calculation result is kept to three decimal places, and the typical value distribution interval is 0.35°-3.20°.

[0084] The feasibility filtering performs a hierarchical filtering strategy. The first layer filtering sets a hard boundary according to the verticality error tolerance threshold (±1.5°), directly eliminating solutions with deviation angles greater than 1.5°. The second layer filtering adds a safety margin of 0.2°, excluding boundary solutions with deviation angles of 1.3°-1.5°. The third layer filtering combines engineering experience rules: when the cooling fluid flow is less than 10 L / min and the rotation speed exceeds 400 rpm, the scheme is forced to be abandoned. During the filtering process, constraint conflicts are checked simultaneously: although a certain scheme meets the verticality requirement (1.2°), its feed pressure of 95 MPa exceeds the upper limit of the current drilling rig model of 90 MPa, triggering the activation mechanism of an alternative scheme.

[0085] The data structure of the final output optimal drilling parameter configuration scheme uses a multi-field dictionary format. Each scheme includes a device control parameter group, a geological adaptation parameter group, and a quality evaluation index group. A typical configuration scheme example is: device parameters: rotation speed 380 rpm / feed pressure 70 MPa / cooling fluid flow 22 L / min; geological adaptation: gravel layer compensation coefficient 1.15 / fracture zone avoidance factor 0.7; evaluation index: predicted drilling speed 1.8 m / h / bit wear increment 0.13% / h / verticality deviation 0.85°. The scheme application stage implants a dynamic monitoring interface: if the actual verticality measurement value exceeds the theoretical value by 20% for three consecutive times, the parameters are rolled back to the previous safe configuration point.

[0086] Example 5: Real-time monitoring of dynamic drift is achieved by installing a triaxial digital inclination sensor at the top of the drill pipe. The sensor collects pitch and yaw angle data at a rate of 20 samples per second, with an angular resolution of 0.01 degrees. After the original angle values are filtered by Kalman filter to eliminate high-frequency noise, the vector difference module of the current angle value and the initial reference angle is calculated. The drift threshold is set as a spatial conical constraint: when the drill pipe end displacement exceeds the conical angle of 1.5 degrees and the duration exceeds 3 consecutive sampling periods, the parameter reconfiguration process is triggered. When triggered at a drilling depth of 8.2 meters, the control system records the current timestamp and vibration spectrum feature snapshot, saving it as an abnormal working condition state flag file.

[0087] The geological parameter updating link is based on the depth-stratum mapping database. The current drilling depth coordinate inputs the tomographic information indexing system, which searches for stratum attributes within a range of ±0.3 meters in the vertical direction. The query returns a gravel layer identification code (GRAVEL-7), which is associated with the stratum hardness reference value (HB320) and the compressive strength coefficient (1.15) in the physical property parameter library. The update logic of the stratum hardness compensation coefficient is: when the stratum hardness change gradient exceeds 15 HB per meter is detected, the compensation coefficient takes the weighted sum of the reference value and the gradient change. In this case, the original coefficient of 1.2 is adjusted to 1.5 according to the gradient change of 12 HB, and the updated constraint condition is injected into the boundary constraint set of the real-time optimization engine.

[0088] The parameter re-computation process starts dynamic memory allocation. The optimization engine retains the target fused drilling state feature matrix from the previous calculation and reloads the updated stratum hardness compensation coefficient. The multi-objective optimization function adds an additional term: the compensation coefficient is introduced as a multiplication factor in the drill bit lateral stress constraint equation, which increases the original constraint boundary by 50%. The iterative solver uses the feasible direction method, and the initial iteration point is selected from the neighborhood of the historical optimal solution (rotation speed ±10 rpm, pressure ±5 MPa). The convergence criterion is set to a change rate of the objective function of less than 0.1% for 5 consecutive iterations, and the maximum number of iterations is limited to 15. The generated new parameter configuration scheme includes a speed reduction instruction (350 rpm) and a pressure increase instruction (85 MPa), as well as an additional drill pipe rotation angle compensation value of 0.3 degrees.

[0089] The vibration spectrum data acquisition uses a wideband acceleration sensor group. Three sensors are installed on the upper, middle and lower parts of the drill pipe, with a sampling frequency covering the 0-2000Hz frequency band. After anti-aliasing filtering, the spectrum data is subjected to short-time Fourier transform with a time window of 0.5 seconds, and the frequency domain resolution is set to 1 Hz. The abnormal resonance frequency band identification uses the spectral kurtosis analysis method: calculate the kurtosis coefficient of each 1Hz frequency band, and when the kurtosis value of a certain frequency band (80-90Hz interval) exceeds 3 times the standard deviation of the reference spectrum, it is marked as an abnormal frequency band. The center frequency of the frequency band is calculated using the energy barycenter method, and the result is rounded to two decimal places (84.7Hz).

[0090] Historical failure case matching executes a bidirectional search strategy. Forward search indexes the center frequency as the key to search for resonance records within a ±2 Hz range in the mechanical failure case library; backward search filters relevant cases according to the current working condition label (gravel layer / high speed). The matched failure case outputs three features: the main shaft bearing gap amount (0.15 mm) when resonance occurs, the gearbox temperature mutation gradient (18 ℃ / min), and the lubricating oil viscosity attenuation rate (2.4% / h). The risk level evaluation index is generated by a decision tree model: input parameters such as abnormal frequency band bandwidth (9.3 Hz), energy increase (24 dB), and duration (8 seconds) are weighted by feature importance, and a risk value of 7.8 points (on a 10-point scale) is output.

[0091] Safety constraint injection adopts a hierarchical progressive mechanism. When the risk value exceeds the safety threshold (7 points), the first layer of constraints forces the upper limit of the rotating speed to 300 rpm; the second layer of constraints introduces a feed force attenuation coefficient of 0.7, which is implemented by limiting the flow of the hydraulic system control valve; the third layer of constraints adds a vibration monitoring intensification strategy, shortening the frequency spectrum analysis window to 0.2 seconds. Self-checking instructions are embedded in the parameter configuration scheme in synchronization: a drill pipe modal analysis is started after each constraint injection, and the first three natural frequencies are calculated using the Lanczos algorithm. If the frequency of 84.7 Hz is in the resonance zone (difference from the first natural frequency <5 Hz) under the new parameters, a rotating speed offset instruction of ±5 rpm is added for frequency domain avoidance. All constraint parameters are packaged into safety protocol data packets, which are transmitted to the actuator through the industrial real-time bus with a transmission delay controlled within 5 milliseconds. The protocol contains a conflict detection flag, which activates an artificial decision request signal when the new parameters conflict with the original task priority.

[0092] It should be noted that, in the present text, relational terms such as first and second are used merely to distinguish one entity or action from another, without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprising", "containing" or any other variant thereof are intended to cover non-exclusive inclusion, so that processes, methods, articles or equipment including a series of elements not only include those elements, but also include other elements not explicitly listed or inherent to such processes, methods, articles or equipment.

[0093] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method of parameter acquisition for a hole rod erecting machine, characterized by, The method comprises the following steps: acquiring multi-source sensing data of a target drilling area, wherein the multi-source sensing data comprises geological structure tomography information and real-time operation state parameters of a mechanical power system; performing working condition matching retrieval based on a historical drilling case library to determine a plurality of heterogeneous working condition cluster sets, and performing feature set extraction by traversing the plurality of heterogeneous working condition cluster sets to obtain a plurality of drilling feature peak values and a plurality of drilling feature diffusion ranges; taking the plurality of drilling feature diffusion ranges as analysis scales of a plurality of feature extraction network layers to perform time sequence feature slicing processing on the multi-source sensing data by the plurality of feature extraction network layers to generate a plurality of drilling state feature sequence sets; performing cross-scale feature interaction fusion on the plurality of drilling state feature sequence sets to generate a target fusion drilling state feature matrix; generating an optimal drilling parameter configuration scheme according to the target fusion drilling state feature matrix and a preset drilling parameter constraint condition; the step of generating the optimal drilling parameter configuration scheme according to the target fusion drilling state feature matrix and the preset drilling parameter constraint condition comprises: analyzing semantic dependency relationships in drilling task instructions to extract drilling depth priority coefficients and a verticality error tolerance threshold; mapping the drilling depth priority coefficients to parameter optimization weight factors, and combining the target fusion drilling state feature matrix to construct a multi-objective optimization function; performing Pareto frontier search on a rotational speed parameter range, a feed pressure threshold and a cooling liquid flow interval by using a non-dominated sorting strategy to generate a candidate parameter configuration set; performing feasibility filtering on the candidate parameter configuration set according to the verticality error tolerance threshold to output the optimal drilling parameter configuration scheme.

2. The parameter acquisition method of a hole making stand according to claim 1, wherein the step of performing working condition matching retrieval based on the historical drilling case library to determine the plurality of heterogeneous working condition cluster sets comprises: extracting geological structure feature vectors and mechanical power system state feature vectors from the multi-source sensing data; calculating the Euclidean distance between the geological structure feature vectors and the case geological features in the historical drilling case library to screen a candidate case subset satisfying a similarity threshold; performing energy spectrum density distribution analysis on the mechanical power system state feature vectors in the candidate case subset to generate a working condition clustering center point set; dividing the historical drilling case library into a plurality of heterogeneous working condition cluster sets containing geological attribute labels and power state labels according to the working condition clustering center point set.

3. The method of claim 2, wherein, the step of performing feature set extraction by traversing the plurality of heterogeneous working condition cluster sets to obtain a plurality of drilling feature peak values and a plurality of drilling feature diffusion ranges comprises: extracting a drilling resistance instantaneous change amount set from the plurality of heterogeneous working condition cluster sets, and performing gradient mutation point detection on the drilling resistance instantaneous change amount set; identifying the spatiotemporal coordinates corresponding to the gradient mutation points to generate a drilling feature peak value spatiotemporal distribution map; taking the drilling feature peak value spatiotemporal distribution map as an index to perform feature correlation diffusion analysis on the plurality of heterogeneous working condition cluster sets, calculating the average influence radius of each gradient mutation point, and generating the plurality of drilling feature diffusion ranges.

4. The parameter acquisition method of a hole making stand according to claim 3, wherein The time sequence feature slicing processing of the multi-source sensing data by the multiple feature extraction network layers generates a plurality of borehole state feature sequence sets, including: According to the sliding window size parameter set by the plurality of borehole feature diffusion ranges, the multi-source sensing data is time-dimensionally segmented and sampled; The segmented and sampled data is input into a parallel convolution branch network to extract rock stratum impedance phase gradient features at different time scales; The rock stratum impedance phase gradient features are time-aligned and spliced to generate a plurality of borehole state feature sequence sets containing vibration spectrum features and drill bit wear state features.

5. The method of claim 4, wherein, The cross-scale feature interaction fusion of the plurality of borehole state feature sequence sets generates a target fusion borehole state feature matrix, including: The first borehole state feature sequence set and the second borehole state feature sequence set are selected to perform feature similarity measurement to generate a feature correlation weight coefficient matrix; According to the feature correlation weight coefficient matrix, the second borehole state feature sequence set is weighted and convoluted to generate a primary fusion feature tensor; The primary fusion feature tensor and the third borehole state feature sequence set are cross-scale feature-interacted, and the feature interaction operation is iteratively executed until the fusion of all borehole state feature sequence sets is completed, and the target fusion borehole state feature matrix is output.

6. The method of claim 1, wherein, Further comprising: Real-time monitoring of dynamic deviation during borehole execution, triggering a parameter reconfiguration process when the drill pipe attitude angle deviation exceeds a preset threshold; According to the mapping relationship between the current borehole depth coordinate and the geological structure tomography scanning information, the rock stratum hardness compensation coefficient in the borehole parameter constraint condition is updated; Based on the updated rock stratum hardness compensation coefficient, the multi-objective optimization function is recalculated to generate a dynamically adjusted optimal borehole parameter configuration scheme.

7. The method of claim 6, wherein, The triggering parameter reconfiguration process includes: Collecting drill pipe vibration modal spectrum data, identifying abnormal resonance frequency bands of the drill pipe vibration modal spectrum data; According to the abnormal resonance frequency band center frequency matching the mechanical resonance features in the historical fault cases, a risk level evaluation index is generated; When the risk level evaluation index exceeds a safety threshold, a rotational speed frequency reduction constraint and a feed force attenuation coefficient are injected into the optimal borehole parameter configuration scheme.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by a processor to implement the steps of the method of any one of claims 1 to 7.

9. A parameter acquisition apparatus of a hole drilling stander, characterized by comprising: A processor and a memory are included, the memory stores a computer program, and the processor executes the computer program to implement the steps of the method of any one of claims 1 to 7. A processor and a memory are included, the memory stores a computer program, and the processor executes the computer program to implement the steps of the method of any one of claims 1 to 7.

Images