A correction method and system of a punching positioning strategy based on force feedback

Abstract

The application discloses a kind of based on force feedback's punch positioning strategy's correction method and system, it is related to intelligent drilling control technical field, including, target punch path is discrete into multiple candidate punch points, constructs graph structure prediction model, generates each point theoretical stress value;Real-time force feedback data is collected in actual drilling process, calculates stress residual, constructs structure residual field and identifies abnormal area;Based on residual distribution characteristics limit path adjustment range, and with target position as guide executes path backtracking reasoning, generates attitude correction path;By controlling equipment posture adjustment, realize the continuous correction of punch positioning.The method has the advantages of strong adaptability, high attitude control precision, structure abnormal response timely, etc., can effectively improve the stability and accuracy of punch operation.

Description

Technical Field

[0001] This invention relates to the field of intelligent drilling control technology, and in particular to a correction method and system for drilling positioning strategies based on force feedback. Background Technology

[0002] With the continuous improvement of automation and intelligence in underground engineering, the demand for force feedback-based intelligent drilling equipment in complex geological environments is increasing. Especially in scenarios such as shield tunneling, rock mass support, and tunnel drilling and blasting, the spatial accuracy of the drilling path directly affects the effectiveness of subsequent anchor placement and structural stability. However, limited by the uncertainties of on-site geology and the interference of real-time disturbance factors, traditional path planning methods typically rely on static three-dimensional path design and regularized stepping strategies, failing to achieve dynamic identification and correction of actual drilling errors. This leads to decreased positioning accuracy, mechanical response deviation, and even drilling failure.

[0003] In existing technologies, some methods attempt to achieve attitude feedback control through position sensors or visual positioning, but these methods cannot accurately reflect the true stress state at the borehole point and are ill-suited to handling nonlinear structural disturbances caused by geological changes along continuous paths. Furthermore, existing force feedback control methods are mostly based on single-point responses, lacking the ability to model the overall structural stability of the path and the evolution of continuous errors graphically. This prevents the inference of subsequent structural variation trends from current errors, limiting the intelligence and continuity of attitude correction. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a correction method and system for a drilling positioning strategy based on force feedback, which solves the problem that existing drilling positioning methods cannot achieve real-time identification of path errors and continuous attitude correction based on force feedback information.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a correction method and system for a drilling positioning strategy based on force feedback, comprising: acquiring a target drilling path and discretizing the drilling path into multiple candidate drilling points;

[0008] A graph structure prediction model is constructed based on all candidate punch points, and the theoretical force values ​​of the candidate punch points are generated based on the graph structure prediction model.

[0009] In the drilling process, the candidate drilling point set is selected as the current drilling point with the smallest spatial distance from the current position, the real-time force feedback data of the current drilling point is collected, the real-time force feedback data is compared with the theoretical force value, the force residual of the current drilling point is obtained, and the structural residual field is constructed based on the force residual;

[0010] According to the structural residual field, the positioning deviation region of the current drilling point is determined, and based on the residual distribution characteristics in the positioning deviation region, the path feasible adjustment range of the current drilling point is limited;

[0011] In the path feasible adjustment range, the path backtracking reasoning is performed with the target drilling position as the guide to generate the posture correction path;

[0012] The drilling device is controlled to adjust the drilling posture according to the posture correction path, and the real-time force feedback data after the posture adjustment is taken as the input data of the next candidate drilling point, the drilling prediction model and the posture correction path reasoning condition are updated, and the continuous correction of the positioning strategy is realized.

[0013] As a preferred scheme of the drilling positioning strategy correction method based on force feedback, wherein: the target drilling path includes a preset target drilling path and a drilling device parameter;

[0014] The target drilling path includes the initial designed drilling starting point and ending point three-dimensional coordinates, the spatial form of the planned drilling path, the geological region information code of the path, and the prior partition information of the target path; the drilling device parameters include the minimum operation resolution, the single drilling effective length, the minimum drilling depth, and the drilling bit diameter and discrete maximum spacing;

[0015] Discretizing the target drilling path into a plurality of candidate drilling points includes: constructing a path expression model based on the target drilling path and the device parameters, combining the geological region code information, calculating the structural stability score of each candidate drilling point; according to the change of the structural stability score between adjacent drilling points, the path segment structure variation degree is determined and the path discrete step length is adjusted; the path point set discretized by the step length is constructed into a candidate drilling point set.

[0016] As a preferred scheme of the correction method of the force feedback based perforation positioning strategy, wherein: the construction of the graph structure prediction model includes constructing the candidate perforation points as nodes in the graph structure, and the input features of each node include the three-dimensional spatial coordinates of the node, the structural stability score, and the historical disturbance feature; based on the measured force feedback data and the theoretical stress value collected in the historical drilling process, the residual set of the node at multiple historical time points is calculated, and the residual standard deviation and the measured force feedback variance are respectively calculated, and the residual standard deviation and the measured variance are combined by weighting to form the historical disturbance feature; the edges in the graph structure are established according to the difference of the structural stability scores between the candidate perforation points, and the weights of the edges are calculated based on the difference of the structural stability scores by an exponential function; the multi-layer structure feature aggregation and propagation operation is performed by using the graph neural network to obtain the node embedding vector of the candidate perforation point, and the node embedding vector is mapped to the corresponding theoretical stress value.

[0017] As a preferred scheme of the correction method of the force feedback based perforation positioning strategy, wherein: the construction of the structure residual field includes collecting real-time force feedback data of the current candidate perforation point during the perforation process, and comparing the real-time force feedback data with the corresponding theoretical stress value to obtain the structural residual value of the candidate perforation point; the structural residual value is mapped to the graph structure node corresponding to the candidate perforation point to generate a structure residual graph containing node residual attributes and graph topology relationships, and residual propagation adjustment is performed based on the structure residual graph to perform aggregation optimization of the graph structure and identify abnormal areas in the path.

[0018] As a preferred scheme of the correction method of the force feedback based perforation positioning strategy, wherein: the residual propagation adjustment includes calculating the residual gradient based on the difference of the structural residual values of adjacent nodes in the graph structure, and adjusting the propagation weight of the edge according to the residual gradient; the graph structure aggregation optimization includes constructing an optimization objective function of residual aggregation according to the node residual value and the propagation weight of the edge, and updating the graph structure based on the optimization objective function to enhance the aggregation expression ability of the structural residual value in the continuous path segment; the initial value of the propagation weight of the edge is set according to the distance between the nodes and the difference of the structural stability scores, and the residual gradient is calculated based on the difference of the structural residual values between adjacent nodes during the propagation process; when the residual gradient is greater than a set connectivity threshold, the propagation weight of the corresponding edge is reduced; when the residual gradient is less than the set connectivity threshold, the propagation weight of the corresponding edge is increased;

[0019] The abnormal area extraction includes: identifying high residual nodes whose node residual values exceed a residual intensity threshold according to a preset residual threshold, and calculating a propagation connectivity of each high residual node, the propagation connectivity being a sum of residual propagation weights of all edges directly connected to the high residual node; when the propagation connectivity is greater than a set connectivity lower limit, determining that the high residual node is a component of an abnormal residual cluster; each abnormal residual cluster includes a node index set, a path segment number, a spatial center position, and an average residual value.

[0020] As a preferred solution of the correction method of the force feedback-based punch positioning strategy, wherein: the determination of the positioning deviation area of the current punch point includes identifying a corresponding graph structure node of the candidate punch point in the structure residual graph, and determining a path segment covered by an abnormal residual cluster to which the graph structure node belongs as the positioning deviation area when the graph structure node belongs to the abnormal residual cluster identified in the structure residual field.

[0021] The limitation of the path feasible adjustment range of the current punch point includes: expanding forward and backward respectively around the graph structure node, and calculating the structure residual values and residual propagation weights between adjacent nodes; when the structure residual values of the adjacent nodes satisfy a decreasing trend and the propagation weights of the corresponding edges are greater than a propagation weight lower limit, continuously selecting graph structure nodes that satisfy the conditions from the graph structure node to form a path feasible adjustment range.

[0022] As a preferred solution of the correction method of the force feedback-based punch positioning strategy, wherein: the generation of the posture correction path includes: obtaining a three-dimensional coordinate of the target punch point, and constructing a guide direction vector according to the spatial position relationship between the graph structure node and the target punch point; a path tangent vector is a unit direction vector generated based on the coordinate difference between the current node and the adjacent node; a path posture deviation angle is determined according to the included angle between the path tangent vector and the guide direction vector; a path correction generation value is calculated based on the structure residual value, the average value of the residual propagation weights of the adjacent edges, and the posture deviation angle.

[0023] The construction of the posture correction path includes: selecting a node sequence with the minimum path correction generation value from the current graph structure node along the guide direction in the path feasible adjustment range.

[0024] In a second aspect, the present application provides a correction system for a force feedback-based punch positioning strategy, including: a path modeling module for obtaining a target punch path and discretizing the path into a plurality of candidate punch points in combination with device parameters;

[0025] A stress prediction module is configured to construct a graph structure prediction model and generate a theoretical stress value of the candidate punch point.

[0026] a feedback processing module, configured to collect real-time force feedback data in the drilling process, and compare the real-time force feedback data with the theoretical force value, to obtain a structural residual;

[0027] a residual analysis module, configured to construct a structural residual field, identify a positioning deviation area, and limit a path feasible adjustment range;

[0028] a path correction module, configured to perform path backtracking reasoning in the adjustment range, and generate a posture correction path guided by a target drilling position;

[0029] a posture control module, configured to adjust a drilling posture according to the posture correction path, and use the adjusted feedback data to update subsequent prediction and path correction conditions, to realize continuous correction of the positioning strategy.

[0030] In a third aspect, the present application provides a computer device, comprising a memory and a processor, and the memory stores a computer program, wherein the computer program is executed by the processor to implement any step of the correction method of the drilling positioning strategy based on force feedback according to the first aspect of the present application.

[0031] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, wherein the computer program is executed by a processor to implement any step of the correction method of the drilling positioning strategy based on force feedback according to the first aspect of the present application.

[0032] The present application has the following beneficial effects: the theoretical force value of the candidate drilling point is generated by constructing a graph structure prediction model, to realize pre-modeling of the path structure response; in the drilling process, real-time force feedback is collected and a structural residual field is constructed, to dynamically identify the force deviation and abnormal area; based on the abnormal residual cluster, the drilling positioning deviation area is extracted, and the path feasible adjustment range is limited in combination with the residual distribution trend, to realize constraint control of the positioning accuracy; in the path feasible adjustment range, path backtracking reasoning is performed in combination with the target drilling direction, to generate a posture correction path, to ensure continuous transition of the drilling direction; the correction result is fed back to the subsequent prediction and deviation correction process, to form a closed-loop optimization mechanism, to realize collaborative optimization of path accurate modeling, force residual identification and continuous correction of the drilling posture, and to improve the positioning stability and drilling accuracy in a complex structure environment. BRIEF DESCRIPTION OF DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0034] Figure 1A flowchart of a correction method for a force-feedback-based punch positioning strategy. DETAILED DESCRIPTION

[0035] In order to make the above objectives, features and advantages of the present application more clear and comprehensible, specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0036] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details given herein. In other instances, well-known methods have not been described in detail in order to avoid unnecessarily obscuring the present application. Therefore, the specific embodiments given herein are not to be interpreted as limiting the scope of the application.

[0037] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is independent or alternative to other embodiments.

[0038] Reference Signs List Figure 1 For one embodiment of the present application, the embodiment provides a correction method and system for a force-feedback-based punch positioning strategy, including the following steps:

[0039] S1: Obtain a target punch path and discretize the punch path into multiple candidate punch points.

[0040] The system obtains a preset target punch path and drilling equipment parameters. The target punch path is generated in the punch design stage and includes the three-dimensional coordinates of the initial design drilling starting point and ending point, the spatial form of the planned punch path, the geological region information code of the path, the advance sequence or prior partition information of the target path (such as section division: near zone, middle zone, far zone, etc.). The geological region code information includes: lithology characteristic code, fault and structure information, whether located near fault zone or fold structure, local abnormal stress region identification, structure direction code, historical disturbance record, residual history of historical punch position, force feedback anomaly label of adjacent region, speed fluctuation or blocked region mark in previous drilling, stratigraphic partition and depth level, stratigraphic number, depth code of the current punch point, section type. The lithology characteristic code includes lithology category code, rock density, rock integrity coefficient and rock layer stiffness change rate.

[0041] The drilling equipment parameters include: minimum operating resolution, single drilling effective length, minimum drilling depth, and drill bit diameter and discrete maximum spacing. In order to facilitate the processing and control of continuous path execution, the punch path is first uniformly established by a parameterized expression method to establish a path expression model, and the path is expressed as a function form of path length parameter wherein , is the total length of the path.

[0042] According to the control resolution of the drilling device, the minimum controllable step length, and the structural distribution changes existing in the geological conditions, the drilling path is discretized to form a set of candidate drilling points with spatial explicitness and sequentiality. A dynamic step length discretization strategy based on the structural variation rate is performed on the drilling path. By calling the geological region information code corresponding to the path paragraph and further extracting the fine-grained lithology code, the dynamic adjustment of the path discretization step length is driven.

[0043] The structural stability variation rate between adjacent points on the path is calculated:

[0044]

[0045]

[0046]

[0047]

[0048]

[0049]

[0050] wherein, represents the structural stability variation rate of the segment on the path, represents the candidate drilling point on the path, represents the candidate drilling point on the path; represents the structural stability score at point , represents the structural stability score at point ; is the lithology code score of the point, is the wave velocity gradient score, is the historical residual anomaly score, is the corresponding weight coefficient, satisfying . The score results can be used to dynamically adjust the path discretization density and the subsequent attitude correction sensitivity. represents the index. Lithology category code (e.g., mudstone = 1, sandstone = 2, limestone = 3…); represents the rock density; represents the rock integrity coefficient (0-1), the lower the value, the more broken; represents the wave velocity gradient change value, reflecting the stiffness change rate of the stratum at the point. is normalized weight; represents the historical residual sequence; represents the measured force feedback; represents the predicted theoretical value; represents the disturbance adjustment coefficient; represents the wave velocity value of the stratum at the point . represents the wave velocity value of the stratum at the point .

[0051] According to the structural stability change rate, the discrete step length of each path segment is dynamically adjusted, and the step length calculation method is , wherein is the step length attenuation coefficient, is the basic sampling interval. In the area with severe structural changes, automatic dense sampling is realized, and in the area with homogeneous and stable structure, the sampling interval is automatically enlarged.

[0052] Each candidate perforation point contains three-dimensional coordinate information of the candidate perforation point , path segment number, path tangent vector information, and index number in the path .

[0053] After the discrete processing is completed, the obtained point set is arranged as a candidate perforation point set , which is the basic input for the subsequent steps. Each perforation point will be used for stiffness prediction modeling and theoretical stress value calculation, and will be used for feedback data comparison and intelligent correction of path posture in the actual perforation process; wherein, represents the total number of candidate perforation points obtained after the target perforation path is discretized.

[0054] By introducing the path expression model and the dynamic step length discretization strategy, the high matching of the perforation path and the drilling equipment parameters is realized, and the path modeling process is more in line with the actual control resolution and physical execution ability of the drilling device. Compared with the traditional fixed step length path discretization method, this method can dynamically adjust the path sampling density according to the complexity of the geological conditions where the path is located, effectively avoid generating redundant data in the structure homogeneous interval, and also can improve the response ability of posture adjustment in the structure variation area.

[0055] The structural stability change rate index is constructed by using the geological region code and fine-grained lithology score corresponding to the path segment, and the wave velocity gradient and historical residual anomaly score are combined for weighted fusion, which can reflect the local structural disturbance characteristics of the path. The scoring results are not only used to guide the non-uniform discrete process of the path, but also provide structural prior for subsequent graph structure modeling and path stability evaluation, improving the adaptability of the entire correction process to complex geological scenes.

[0056] Further, the discrete precision is jointly controlled by the step attenuation factor and the basic sampling interval to establish a structure-dominated path segmentation mechanism, ensuring denser sampling and finer control in sensitive geological disturbance areas, while reducing the redundant calculation burden in low-risk areas. This approach enables the candidate punch point set to have three characteristics: spatial order, structural significance, and operational controllability, effectively improving the expression ability of subsequent graph neural network models in node feature processing and graph structure stability modeling.

[0057] Furthermore, each candidate punch point carries complete three-dimensional coordinates, path segment numbers, and path tangent vector information in the output, not only improving the completeness of path expression, but also providing directional reference for the construction of the attitude correction path. This mechanism establishes a direct mapping relationship between path continuity and attitude guidance, breaking the limitations of traditional punch paths that are only based on geometric coordinates and do not have attitude reference information.

[0058] S2: Construct a graph structure prediction model based on all candidate punch points, and generate theoretical stress values of the candidate punch points based on the graph structure prediction model.

[0059] After completing the dynamic discretization of the punch path in step S1, the system obtains a point set composed of multiple candidate punch points as input for subsequent stiffness prediction modeling.

[0060] To intelligently predict the structural response state of each candidate point in the punch path, a graph structure prediction model based on graph neural network is constructed, which integrates path structure topology and node structure information to realize multi-point coupling prediction. Specifically, the system models the candidate point set as a structure graph .

[0061] Each node corresponds to a candidate punch point , and includes the following input feature items: spatial coordinate information , used to maintain path segment spatial continuity; structural stability score , from the structural stability comprehensive evaluation in step S1, which integrates lithology, wave velocity, and historical anomaly features; historical disturbance features , describing the response volatility of the point in historical drilling.

[0062]

[0063] wherein, represents a set of historical residuals; represents a perturbation adjustment weight; represents the greater, the higher the prediction uncertainty, for adjusting the confidence of the model for this point. The final node feature vector is constructed as:

[0064]

[0065] Fusion of "structure-space-stability" three kinds of information, for driving the stiffness propagation and heterogeneous response modeling between nodes in the graph structure.

[0066] Edge represents the topological connection relationship between adjacent nodes in the path, and the weight of the edge in the graph is calculated based on the difference value of the structure stability score between candidate punch points. The smaller the difference value of the structure stability score between any two adjacent nodes 、 , the more similar their structure states and the stronger the continuity, and the higher the weight of the corresponding edge. On the contrary, if the difference between the structure stability scores of the two is large, it means that there is a structural mutation or stability jump in this place, and the system will significantly weaken the transmission ability of the edge and reduce its influence in the feature propagation process. The weight calculation formula of the edge is represented as:

[0067]

[0068] wherein, represents the weight of the edge between node and node in the graph structure; represents an exponential function; is the weight sensitivity control coefficient of the edge; represents the difference value of the structure stability score between candidate punch points and The structure stability score is obtained by calculation in step S1, which reflects the comprehensive stability degree of multiple source structure information such as lithology, wave velocity, and historical residual at the point.

[0069] A three-layer graph convolutional neural network (GCN) is used as the core modeling structure, and the functions of each layer are as follows:

[0070] Input layer: receives the initial features of each node , and performs embedding initialization according to the graph structure.

[0071] Graph convolution layer: each layer performs a structure-aware aggregation process, which is in the form of:

[0072]

[0073] wherein, denotes the adjacency matrix plus self-loop; denotes the corresponding degree matrix; denotes the first layer feature matrix; denotes the first layer weight; is an activation function such as ReLU.

[0074] Output layer: final node feature will be mapped to the theoretical axial force prediction value:

[0075]

[0076] wherein, denotes the theoretical axial force prediction value of node ; denotes the feature vector output by the graph neural network at the layer (i.e. the last layer) of node ; denotes the weight vector of the output layer (transposed form); denotes the bias term of the output layer.

[0077] To ensure that the model output is consistent with the true physical response of the path, a path continuity regularization mechanism is introduced to introduce a smoothing constraint on the output values between adjacent prediction points:

[0078]

[0079] wherein, denotes the path continuity regularization term; denotes that there is a graph structure connection edge between node pair and ; denotes the structural weight of the edge between node and node , which is defined according to the difference in structural stability score; denotes the predicted axial force value of node ; denotes the predicted axial force value of node ; denotes the square of the difference between the predicted values, which is used to measure the stiffness prediction continuity of the path segment.

[0080] The final training loss function is defined as:

[0081]

[0082] wherein, denotes the standard supervised loss; denotes a regularization term adjustment factor; denotes a final loss function of model training; denotes a node actual axial force value.

[0083] The introduction of the graph structure prediction model can uniformly model the spatial position, structural stability and historical disturbance characteristics of each candidate point in the perforation path. By constructing a node-edge network containing spatial continuity and structural coupling relationship, the limitations of traditional point-to-point stiffness evaluation methods in describing the overall coupling relationship of the path structure are broken through, and the comprehensiveness and accuracy of structural response prediction are significantly improved.

[0084] Through multi-source fusion of node features, not only the spatial geometric characteristics of the path are retained, but also geological related information such as lithology, wave velocity and historical anomalies are introduced, and then a structure state vector with strong expression ability is constructed, enhancing the identification ability of the prediction model to the combined action of multiple factors in complex geological environment. Compared with the traditional fitting method based on single point input, the potential structural continuity and difference between path segments can be more fully captured.

[0085] Further, in the edge weight calculation, an exponential decay mechanism is constructed by the difference in structural stability score, so that the model can adaptively weaken the influence of the mutation area and enhance the influence of the stable segment during feature propagation, realizing the targeted propagation of structural features. It effectively solves the problem of prediction jitter or misjudgment caused by local mutations in path modeling, and helps to improve the stiffness prediction accuracy of low disturbance areas in the path.

[0086] Further, a three-layer graph convolution network is used and a path continuity regularization term is superimposed, so that the model can maintain the sensitivity of local structural differences while the overall output presents physical continuity, avoiding unnatural mutations in the connection between path segments, and ensuring that the theoretical axial force value is consistent with the real path stiffness performance. The output node embedding vector is mapped to the theoretical axial force value, which can be directly used for subsequent residual comparison and attitude reasoning, and has good interface consistency and engineering adaptability.

[0087] S3: In the perforation process, the candidate perforation point with the smallest spatial distance from the current position is selected from the candidate perforation point set as the current perforation point; real-time force feedback data of the current perforation point is collected, and the real-time force feedback data is compared with the theoretical force value to obtain the force residual of the current perforation point, and a structure residual field is constructed based on the force residual.

[0088] In the drilling process, the system selects the graph structure node with the smallest spatial distance from the candidate drilling point set as the current drilling point according to the spatial distance relationship between the current position of the drilling device and the discrete point set of the target path; the spatial distance is calculated based on the Euclidean distance, ensuring that the current drilling point is highly matched with the actual position of the drilling head, and ensuring the accuracy of the comparison between the subsequent force feedback data and the theoretical stress value.

[0089] In the actual drilling process, the system performs real-time feedback collection and structural residual analysis on each candidate drilling point based on the theoretical stress value prediction result generated in step S2. By mapping the residual result to the structure graph, combining the updating mechanism of nodes and edges in the graph neural network, a structural residual graph domain model is formed, realizing intelligent identification and spatial aggregation of abnormal response areas in the path, and providing support for subsequent pose correction path generation.

[0090] By mapping the residual result to the structure graph, combining the updating mechanism of nodes and edges in the graph neural network, a structural residual graph domain model is formed, realizing intelligent identification and spatial aggregation of abnormal response areas in the path, and providing support for subsequent pose correction path generation.

[0091] It should be emphasized that the core data basis of this residual graph is derived from the dynamic comparison between the force feedback response of the drilling device and the theoretical stress value in the model. The structural residual value of each node essentially reflects the disturbance effect of the real lithology condition on the drilling stress at the current position, and the system inversely deduces the occurrence position of lithology inconsistency or structural disturbance through this physical deviation, and realizes its aggregation expression in the path structure through the graph structure propagation mechanism. Therefore, the residual field is not isolated from the lithology, but integrates the lithology score in the geological code, the historical disturbance information and the real-time force feedback response, so as to realize the indirect mapping of lithology characteristics in the graph neural framework.

[0092] When the drilling device advances to the candidate drilling point , the system collects the measured axial stress value of the current point through the force sensor. At the same time, the corresponding point theoretical stress prediction value output by S2 is called to calculate the stress residual:

[0093]

[0094] Wherein, represents the structural stress residual value of the th candidate drilling point, which represents the deviation between the actual stress and the theoretical predicted stress of the point; represents the measured axial stress value of the th drilling point collected through the sensor; represents the theoretical stress prediction value of the th drilling point output by the graph neural network model.

[0095] The residual values ​​of all punched points are summarized into a residual vector sequence:

[0096]

[0097] in, This represents the set of residual values ​​for all punched points in the current path; This indicates the number of candidate drilling points that have completed drilling and force feedback collection.

[0098] To ensure that the residual information in the subsequent graph structure remains consistent with the structural properties of the path points, the system will process each candidate punch point in the path. With nodes in the structure diagram Establish a one-to-one correspondence. Node The associated structural residual values This means that it is at the path point. The force deviation between the actual force feedback collected at the site and the force prediction of the model is completely consistent in terms of numbering, spatial location and structural meaning, and can be used as the input basis for subsequent residual map modeling and anomaly identification.

[0099] Each punched point Mapping back to the structure graph Construct a structural residual graph based on the node attributes:

[0100]

[0101] node Including coordinates Structural stability score Path number and the current residual value .in, Indicates the first The graph nodes corresponding to each candidate punch point. Represents a node The corresponding three-dimensional spatial coordinates.

[0102] To achieve dynamic weight control of edges in the abnormal response segment of the graph, the system defines the residual gradient:

[0103]

[0104] in, Represents a node and The residual gradient value between the two points represents the degree of change in the force deviation between them. This represents the structural stress residual value of the j-th candidate punching point.

[0105] And adjust the weight of the residual propagation edge in the structure diagram:

[0106]

[0107] wherein, represents the residual propagation weight of the edge between the node pair . represents the residual gradient sensitivity coefficient.

[0108] In the residual graph construction process, the system introduces an aggregated energy constraint function in the graph structure, so that the residual value is naturally focused in the structure continuous region, and the structure abnormal cluster is highlighted, which is defined as follows Graph residual consistency regularization term:

[0109]

[0110] wherein, represents the residual consistency regularization term in the graph structure.

[0111] Combine the residual amplitude term to form the residual graph optimization target:

[0112]

[0113] wherein, represents the optimization target function of the structure residual graph, represents the regularization term adjustment factor, which controls the weight of the aggregated regularization term in the total optimization target.

[0114] The graph structure will automatically adjust the residual transmission path between nodes and the convergent structure according to this target function, automatically adjust the information flow path between nodes, so that the high residual nodes are aggregated into an abnormal cluster structure with spatial continuity and structural consistency. Based on the optimized residual graph structure, the system automatically extracts the abnormal area in the response set in the path segment, which is defined as the residual clustering cluster set:

[0115]

[0116] wherein, represents the residual clustering cluster set; represents the th clustering cluster; represents the th node in the graph structure, from the node set represents the high residual point when the residual value of the node is greater than the preset residual intensity threshold . is the propagation connectivity, represents the high propagation connectivity when the sum of the propagation connectivity of the edge connected to the node exceeds the connectivity lower limit .

[0117] After the system performs force feedback collection, residual calculation, graph structure mapping, and graph domain aggregation processing, the structural residual field is constructed, and it is taken as the input basis of the path attitude correction reasoning module. The structural residual field includes residual graph node attributes, edge residual propagation weights, and an abnormal residual cluster set.

[0118] Residual graph node attributes: record the structural residual value of each candidate punch point node obtained by comparing real-time feedback and model prediction.

[0119] Edge residual propagation weight: the edge propagation coefficient is constructed according to the adjacent node residual gradient and the edge weight joint structure stability score , used to constrain the propagation ability of information in the path paragraph in the graph.

[0120] Abnormal residual cluster set: high response area automatically aggregated during graph structure optimization, each abnormal residual cluster contains node range, path number index segment, spatial center point, and average residual intensity attribute fields. The structural residual field provides a structural expression based on feedback deviation distribution for the subsequent path adjustment step, supporting the starting point positioning, range judgment, and direction generation of the attitude correction path.

[0121] The mechanism of selecting the current punch point based on the spatial nearest neighbor principle is introduced, combined with the Euclidean distance calculation method, which effectively ensures the high matching between the current position of the punch device and the candidate point, so that the comparison between the subsequent real-time force feedback and the theoretical stress value has consistency and accuracy guarantee. Compared with the traditional method relying on fixed step or sequential advancement, this method improves the positioning flexibility and response accuracy of the system under complex spatial path conditions.

[0122] By comparing the real-time collected force feedback data with the theoretical stress value output by the graph structure prediction model point by point, and constructing a structural residual vector sequence in real time, the system not only accurately quantifies the mechanical deviation of the punch point, but also gradually accumulates spatial response characteristics during path advancement, significantly enhancing the ability to identify local abnormal segments of the path. Traditional methods rely more on global average or single-point anomaly judgment, which is difficult to reflect local details and path continuity.

[0123] Further, in the residual graph modeling process, the residual value is deeply bound with the node attribute, and the residual gradient calculation and edge propagation weight adjustment mechanism are introduced, so that the information propagation in the graph structure has dynamic adjustment ability. This design can effectively suppress the diffusion of abnormal disturbances and enhance the aggregation trend of high residual areas, with stronger abnormal focusing ability. Compared with traditional static graph structure modeling methods, this mechanism has significant advantages in dynamic adjustment of residual information.

[0124] Further, by constructing an optimization objective function through residual consistency regularization and residual amplitude term, the system guides the automatic completion of abnormal region clustering and high residual cluster extraction of the graph structure, forms a structural residual field, and provides a boundary clear and structure reasonable input basis for subsequent attitude correction reasoning. Compared with the traditional punch path correction which relies on artificial threshold setting or empirical local strategy, this method has substantial improvement in the automation of abnormal identification, the continuity of structure expression, and the accuracy of path correction.

[0125] S4: Determine the positioning deviation region of the current punch point according to the structural residual field, and limit the path feasible adjustment range of the current punch point based on the residual distribution characteristics in the positioning deviation region.

[0126] After the structural residual field is constructed, the system identifies whether there is a structural response deviation of the current punch point based on the node residual value and graph structure connection information contained in the structural residual field, and limits the path attitude adjustment range that can be executed at the current position. Specifically, the system first determines the corresponding node of the current candidate punch point in the graph structure, and judges whether the node belongs to any abnormal residual cluster. If the current graph node belongs to a residual cluster , it indicates that there is a stability abnormality or force feedback deviation phenomenon in the path structure at the current position, and the path segment covered by the residual cluster is regarded as the positioning deviation region of the punch point.

[0127] The residual cluster is composed of a group of graph nodes with high residual value and high propagation connectivity, the system extracts the path segment containing the current node inside the cluster, and analyzes the structural disturbance direction of the current path point according to the residual distribution change trend of the nodes in the cluster. The residual distribution change trend specifically includes: expanding forward and backward from the current node, respectively, counting the residual value change of adjacent nodes, and judging whether it satisfies the residual gradient decreasing trend.

[0128] At the same time, the system calculates the residual propagation weight of the edge between the current node and the adjacent node, and judges the residual distribution change trend; when the residual propagation weight of the edge is greater than the set lower limit of the propagation weight, and the residual values between adjacent nodes present a decreasing trend, the system considers that the path direction has structural continuity and error adjustability. The system starts from the current node and continuously filters the nodes that meet the above conditions along the path direction to construct the path feasible adjustment range of the current position. The path feasible adjustment range is expressed in the form of node number range, which represents the path sub-segment that can be adjusted in attitude at the current punch point within the structural allowable range.

[0129] By introducing structural residual clusters as the basis for identifying mechanically abnormal regions in the path, the system achieves accurate identification of drilling positioning deviation regions. This method uses node residual values ​​and graph structure propagation connectivity as core criteria, breaking through the coarse-grained judgment mode of traditional methods that rely solely on single-point thresholds or fixed tolerance ranges, and significantly enhancing the system's discrimination resolution and positioning capability for abnormal regions.

[0130] In determining the path disturbance trend, a complete residual gradient analysis system was constructed by statistically analyzing the distribution changes of residual values ​​in the preceding and following directions of the path. The system can automatically identify the direction of residual decrease and determine whether the structural response has continuity, thereby achieving dynamic identification of path adjustability. This approach differs from traditional methods that rely on static empirical values ​​to define the adjustment range; it can adapt in real time according to the current state of the path, thus offering greater response flexibility and structural adaptability.

[0131] Furthermore, the introduction of propagation connectivity means that the determination of the feasible adjustment range of a path is not only based on the residual magnitude of a single node, but also comprehensively considers the changes in edge weights between adjacent nodes and the overall connectivity of the graph structure. This adjustment boundary construction method based on "structural propagation feasibility" effectively avoids misidentifying locally high residual nodes as overall structural instability, thus improving the accuracy of the attitude adjustment range.

[0132] Furthermore, by continuously screening nodes within the path segment that satisfy the conditions of decreasing residuals and high connectivity of edge weights, a feasible adjustment range for the path is constructed. This provides subsequent path correction strategies with a spatial structural basis and physical feasibility guarantee, which is significantly better than the traditional adjustment segment method that relies on a fixed range sliding window method. It shows significant progress in terms of structural rationality, path adaptability, and execution security.

[0133] S5: Within the feasible adjustment range of the path, perform path backtracking reasoning guided by the target drilling position to generate an attitude correction path.

[0134] After the feasible adjustment range of the path is constructed, the system, based on the structural residual map and path space topology information, performs a path backtracking reasoning process guided by the target drilling point, and outputs an attitude correction path for drilling attitude adjustment. First, the system extracts the three-dimensional coordinates of the target drilling point as a reference position:

[0135]

[0136] in, Represents the three-dimensional spatial coordinates of the target drilling point. These represent the target drilling points in three-dimensional space. , , Coordinate values.

[0137] And in the path feasible adjustment range defined in the graph structure node set, select the graph structure node corresponding to the current candidate punch point , get three-dimensional coordinate information, and construct a guide direction based on the spatial position difference between the coordinates and the target punch point; the guide direction is a unit vector dynamically calculated according to the geometric relationship between the current path position and the target position:

[0138]

[0139] Among them, represents the unit guide direction vector from the candidate punch point to the target punch point ; represents the three-dimensional coordinates of the graph structure node corresponding to the candidate punch point numbered in the path; represents the Euclidean distance between the candidate punch point and the target punch point (i.e. the vector length).

[0140] Calculate the attitude deviation angle of the current node , that is, the included angle between the path tangent vector of the node and the target direction, which is used to evaluate the direction consistency. The path tangent vector is the unit direction vector between the two adjacent candidate punch points in the path, which represents the advancing direction or "tangent direction" of the path at the point.

[0141] On this basis, the system establishes a path cost function:

[0142]

[0143] Among them, represents the path correction cost value of the graph structure node ; respectively represent the weighted coefficients of the three sub-items in the path cost function, satisfying represents the structural residual value (i.e. the difference between the actual force and the theoretical force) of the graph structure node ; represents the average value of the residual propagation edge weight between the node and all its adjacent nodes.

[0144] Based on the path cost function, the backtracking path reasoning is performed, starting from the current candidate punch point node, gradually selecting the graph structure node with the minimum path cost value in the path feasible adjustment range along the guide direction, to construct a continuous node sequence as the attitude correction path. The attitude correction path is a sequence of node numbers:

[0145]

[0146] wherein, represents a pose correction path node sequence; represents a graph structure starting node corresponding to the current candidate puncture point; represents a terminal graph structure node closest to the target puncture direction in the pose correction path; represents the first graph structure node in the pose correction path; represents the last graph structure node in the pose correction path.

[0147] The system extracts the path tangent vector in the above node sequence, constructs a control vector sequence of the pose correction path, and provides it to the pose adjustment control module of the puncture device to complete the continuous correction of the puncture pose.

[0148] By introducing the position of the target puncture point as a dynamic guide direction within the feasible adjustment range of the path, a path backtracking reasoning process based on actual geometric relationships is constructed, effectively avoiding the direction error accumulation problem caused by generating a correction path based on static rules or linear interpolation in traditional pose correction. By calculating the unit direction vector between the candidate puncture point and the target puncture point in real time, the entire correction path is guided to converge to the structure target area, improving the spatial accuracy of path reconstruction.

[0149] In the path node screening process, the angle between the path tangent vector and the guide direction is introduced as a measure of pose deviation, significantly enhancing the control ability of path reconstruction on direction continuity. Compared with the previous selection strategy based on residual strength or path length, this mechanism can ensure that the selected path remains stable and consistent in the advancing direction, thereby improving the physical reasonableness and execution stability of the device pose adjustment process.

[0150] Further, the design of the path correction cost function integrates three key factors: structure residual value, average value of residual propagation weight, and path direction consistency, evaluating the local quality of path correction from multiple dimensions. This composite cost function not only filters out nodes in areas with high local disturbance or structural discontinuity, but also prioritizes path segments with reasonable direction and structural coherence, ensuring that the final generated correction path has strong execution feasibility and structural stability.

[0151] Further, by dynamically calculating the path cost function and backtracking the optimal node sequence, the system constructs a pose correction path and further extracts the path tangent vector as a control vector sequence to provide to the puncture execution module, ensuring operation continuity and trajectory smoothness during pose correction. This path generation method, which is driven by graph structure, guided by direction consistency, and fused with multiple dimensional constraints, has stronger adaptability and structural adaptability compared to traditional one-way backtracking strategies.

[0152] S6: The control drilling device adjusts the drilling pose according to the pose correction path, and uses the real-time force feedback data after the pose adjustment as the input data of the next candidate drilling point, updates the drilling prediction model and the correction path inference condition, and realizes continuous correction of the positioning strategy.

[0153] After the pose correction path is generated, the system calls the path tangent vector sequence corresponding to each node in the path as the pose control reference information of the drilling device, and continuously adjusts the device pose. Based on the advancing direction characteristics represented by the path tangent vector, the control process dynamically corrects the deviation between the current pose of the drill bit and the target pose, ensuring stable advancement along the optimal path direction during the drilling process.

[0154] Specifically, the system maps each graph structure node in the pose correction path to the corresponding three-dimensional coordinates in the path space, and calculates its path tangent vector to form a set of pose control instructions. The control system drives the drilling device to perform segmented advancement and pose fine-tuning using this instruction set, and performs position correction and direction correction at each step to ensure high-precision pose consistency during path adjustment.

[0155] After the drilling device completes the correction path advancement, the system collects real-time force feedback data at the current drilling point, and calculates a new structural residual for the current node based on the measured axial force value and the corresponding theoretical force value. This structural residual is mapped to the graph structure node to update the residual attribute and propagation edge weight in the graph structure, and then iteratively optimize the structural residual field model.

[0156] Based on the updated structural residual field, the system re-executes the abnormal cluster identification, path feasible adjustment range limitation and path backtracking inference process to continuously generate a new pose correction path. This mechanism realizes continuous updating of the positioning strategy driven by force feedback data, ensuring that the drilling device can achieve adaptive pose adjustment and dynamic response optimization in complex structural environments, thereby significantly improving the stability and structural matching accuracy of the drilling path tracking.

[0157] The embodiment also provides a drilling positioning strategy correction system based on force feedback, comprising:

[0158] A path modeling module is configured to obtain a target drilling path and discretize the path into a plurality of candidate drilling points in combination with device parameters.

[0159] A force prediction module is configured to construct a graph structure prediction model to generate theoretical force values for the candidate drilling points.

[0160] A feedback processing module is configured to collect real-time force feedback data during the drilling process and compare it with the theoretical force values to obtain a structural residual.

[0161] A residual error analysis module is configured to construct a structural residual error field, identify a positioning deviation area, and define a path feasible adjustment range.

[0162] A path correction module is configured to perform path backtracking reasoning in the adjustment range and generate a pose correction path guided by a target drilling position.

[0163] A pose control module is configured to adjust a drilling pose according to the pose correction path and use the adjusted feedback data to update subsequent prediction and path correction conditions, so as to realize continuous correction of the positioning strategy.

[0164] The embodiment also provides a computer device suitable for the correction method of the drilling positioning strategy based on force feedback, which includes a memory and a processor.

[0165] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to perform wired or wireless communication with an external terminal. The wireless communication can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. In addition, the input device can be an external keyboard, touchpad or mouse, etc.

[0166] The embodiment also provides a storage medium, which stores a computer program, and the computer program is executed by a processor to implement the correction method of the drilling positioning strategy based on force feedback proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic storage, a flash memory, a magnetic disk, or an optical disk.

[0167] The embodiment also provides a correction method and system of the drilling positioning strategy based on force feedback, and scientific demonstration is performed through a simulation experiment to verify the beneficial effects of the application.

[0168] An experiment is performed in a simulated tunnel drilling engineering scene, the total length of a target path is set to 10.00 meters, the path includes a slightly curved section, and a geological region is a type II hard rock. A numerical control drilling machine provided with a high-precision axial force sensor is used as a drilling device, and device parameters include: a minimum operation resolution is 0.10 meters, a single drilling effective length is 0.20 meters, and a maximum allowed attitude deviation angle is 2.00°. Based on the path length and the device parameters, the path is discretized into 5 candidate drilling points, sequentially numbered as P1 to P5, and a spatial position interval is 2.00 meters.

[0169] Firstly, a target path and device parameters are acquired, a candidate drilling point set is constructed, and a graph structure prediction model is trained based on historical drilling data to generate theoretical force values of each point. The theoretical values are: P1 is 9.70 kN, P2 is 9.60 kN, P3 is 9.50 kN, P4 is 9.40 kN, and P5 is 9.30 kN. Subsequently, real-time measured force feedback values are collected in the drilling process of each candidate point, and are recorded as: P1 is 9.68 kN, P2 is 9.59 kN, P3 is 8.95 kN, P4 is 8.80 kN, and P5 is 8.75 kN. The measured values are compared with the theoretical values, and residual values are calculated. The residual deviations of P3 to P5 are large, and are -0.55 kN, -0.60 kN, and -0.55 kN respectively.

[0170] The system constructs a structure residual graph based on node residuals, identifies P3-P5 region as an abnormal residual cluster, and extracts the spatial distribution and propagation connectivity features of the nodes in the cluster. Further analysis finds that the propagation connectivity between P4 and P5 has significantly decreased, indicating that there is a structural disturbance in this path segment. The system limits the feasible adjustment range of the path to P2 to P5 and constructs a pose correction path guided by the target point P5 to guide the drill bit to adjust its pose in the optimal direction. After the correction path is generated, the control system executes the adjustment instructions and collects the real-time force data of the drilling points again.

[0171] After the pose correction, the real-time force values of the drilling points are significantly improved. After P3 correction, the force is 9.48 kN (before correction, 8.95 kN), after P4 correction, the force is 9.10 kN (before correction, 8.80 kN), and after P5 correction, the force is 9.25 kN (before correction, 8.75 kN). The force deviation after correction is reduced to 0.02 kN, 0.30 kN, and 0.05 kN, respectively, with a residual reduction of more than 80.00%. In addition, the attitude deviation angle of the path is also monitored: before adjustment, the attitude deviation angles of P4 and P5 are 1.85° and 2.10°, respectively, and after adjustment, they are reduced to 1.00° and 0.75°, effectively reducing the risk of attitude deviation.

[0172] In terms of propagation connectivity, the average propagation connectivity between P4 and P5 before correction is 0.42, and after correction, it is increased to 0.87, indicating that the structural connectivity has been significantly restored, and the information transmission path in the graph structure is more stable, and the abnormal area aggregation degree is controlled. Through the convergence value comparison of the aggregation optimization objective function, the value before correction is 1.72, and the value after correction is 0.45, further indicating that the abnormal residual cluster in the graph structure is effectively separated and corrected.

[0173] Through this experiment, it can be clearly seen that the proposed drilling positioning strategy correction method based on force feedback is significantly better than the traditional fixed path drilling method in practical application. The traditional method relies on pre-set paths and manual adjustment, lacks automatic identification and correction mechanism for local structural disturbance or geological anomalies, resulting in serious force mismatch, often causing drill bit deviation, local structure damage, and path error accumulation. In this experiment, the system predicts the model by constructing a graph structure, uses node residuals and propagation connectivity to automatically identify abnormal structure areas, and solves the problem of structural changes that cannot be perceived by manual adjustment.

[0174] Especially in P3 to P5, the structural changes that cannot be perceived by the traditional method are explicitly expressed through the residual field, enabling the system to identify potential deviations in advance and dynamically limit the adjustment range. The backtracking reasoning of the pose correction path not only considers the geometric guidance direction, but also integrates the structure residual and propagation characteristics, making the generated path highly robust and locally continuous, ensuring that the drilling equipment advances along the optimal path.

[0175] In addition, after the attitude correction, not only the single-point force value is improved, but also the deviation of P3 to P5 points is reduced by more than 0.50kN on average, and the overall path structure connectivity and information aggregation are greatly optimized. The propagation connectivity is improved from less than 0.50 to close to 0.90, proving the actual ability of the graph neural network in information fusion and abnormal focusing. Through the continuous feedback collection and path backtracking update in the punching process, the whole strategy forms a closed-loop control process with adaptive ability, providing a new technical solution for high-precision drilling in complex structure scenarios.

[0176] In summary, the method of the present application has obvious technical progress in ensuring the accuracy of path execution, the accuracy of attitude control, the timeliness of structure anomaly perception, etc., and has high practical value and popularization prospect.

[0177] It should be noted that the above examples are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A correction method of a force feedback-based punch positioning strategy, characterized by: The method comprises the following steps: acquiring a target drilling path and discretizing the drilling path into a plurality of candidate drilling points; constructing a graph structure prediction model based on all candidate drilling points and generating theoretical stress values of the candidate drilling points based on the graph structure prediction model; in the drilling process, selecting a candidate drilling point with the smallest spatial distance from the current position as the current drilling point, collecting real-time force feedback data of the current drilling point, comparing the real-time force feedback data with the theoretical stress values, obtaining a stress residual of the current drilling point, and constructing a structural residual field based on the stress residual; the construction of the structural residual field comprises collecting real-time force feedback data of the current candidate drilling point and comparing the real-time force feedback data with the corresponding theoretical stress values to obtain a structural residual value of the candidate drilling point; mapping the structural residual value to a graph structure node corresponding to the candidate drilling point to generate a structural residual graph containing node residual attributes and graph topology relationships, and performing residual propagation adjustment based on the structural residual graph, performing aggregate optimization of the graph structure, and identifying abnormal areas in the path; the residual propagation adjustment comprises calculating a residual gradient based on the structural residual value difference between adjacent nodes in the graph structure, and adjusting the propagation weight of the edge according to the residual gradient; determining a positioning deviation area of the current drilling point according to the structural residual field, and limiting a path feasible adjustment range of the current drilling point based on residual distribution characteristics in the positioning deviation area; performing path backtracking reasoning in the path feasible adjustment range to generate a posture correction path; the generation of the posture correction path comprises acquiring three-dimensional coordinates of the target drilling point and constructing a guide direction vector according to the spatial position relationship between the graph structure node and the target drilling point; a path tangent vector is a unit direction vector generated based on the coordinate difference between the current node and the adjacent node; and a path posture deviation angle is determined according to the included angle between the path tangent vector and the guide direction vector; calculating a path correction generation value based on the structural residual value, the average value of the residual propagation weight of the adjacent edge, and the posture deviation angle; constructing the posture correction path comprises selecting a node sequence with the minimum path correction generation value from the current graph structure node along the guide direction in the path feasible adjustment range; controlling the drilling equipment to adjust the drilling posture according to the posture correction path, taking the real-time force feedback data after the posture adjustment as the input data of the next candidate drilling point, updating the drilling prediction model and the posture correction path reasoning condition, and realizing continuous correction of the positioning strategy.

2. The force feedback based punch location strategy correction method of claim 1, wherein: the acquisition of the target drilling path comprises acquiring a preset target drilling path and drilling equipment parameters; the target drilling path comprises three-dimensional coordinates of an initially designed drilling starting point and ending point, a spatial form of a planned drilling path, geological region information coding of the path, and prior partition information of the target path; the drilling equipment parameters comprise a minimum operating resolution, a single drilling effective length, a minimum drilling depth, and a drilling bit diameter and a discrete maximum spacing; The target perforation path is discretized into a plurality of candidate perforation points, including constructing a path expression model based on the target perforation path and device parameters, combining geological region coding information, and calculating the structural stability score of each candidate perforation point; According to the change of the structural stability score between adjacent perforation points, the structural variation degree of the path segment is determined and the discrete step length of the path is adjusted; the path point set discretized by the step length is constructed into a candidate perforation point set.

3. The force feedback based punch location strategy correction method of claim 2, wherein: The construction of the graph structure prediction model includes constructing the candidate perforation points into nodes in the graph structure, and the input features of each node include the three-dimensional spatial coordinates of the node, the structural stability score and the historical disturbance feature; Based on the measured force feedback data and the theoretical stress value collected in the historical drilling process, the residual set of the node at multiple historical time points is calculated, and the residual standard deviation and the variance of the measured force feedback are respectively calculated, and the residual standard deviation and the measured variance are combined to form the historical disturbance feature; According to the difference of the structural stability score between the candidate perforation points, the edges in the graph structure are established, and the weights of the edges are calculated based on the difference of the structural stability score by an exponential function; The graph neural network is used to perform multi-layer structure feature aggregation and propagation operation to obtain the node embedding vector of the candidate perforation point, and map it to the corresponding theoretical stress value.

4. The force feedback based punch location strategy correction method of claim 3, wherein: The graph structure aggregation optimization includes constructing an optimization objective function of residual aggregation according to the node residual value and the propagation weight of the edge, and updating the graph structure based on the optimization objective function to enhance the aggregation expression ability of the structural residual value in the continuous path segment; the initial value of the propagation weight of the edge is set according to the distance between nodes and the difference of the structural stability score, and the residual gradient is calculated based on the structural residual difference between adjacent nodes in the propagation process, when the residual gradient is greater than the set connectivity threshold, the propagation weight of the corresponding edge is reduced; When the residual gradient is less than the set connectivity threshold, the propagation weight of the corresponding edge is increased; The abnormal region extraction includes: identifying high residual nodes whose node residual values exceed the residual intensity threshold according to the preset residual threshold, and calculating the propagation connectivity of each high residual node, the propagation connectivity being the sum of the residual propagation weights of all edges directly connected to the high residual node; when the propagation connectivity is greater than the set connectivity lower limit, the high residual node is determined to be a component of an abnormal residual cluster; each abnormal residual cluster includes a node index set, a path segment number, a spatial center position and an average residual value.

5. The force feedback based punch location strategy correction method of claim 4, wherein: The determination of the positioning deviation region of the current perforation point includes identifying the corresponding graph structure node of the candidate perforation point in the structural residual graph, and when the graph structure node belongs to the abnormal residual cluster identified in the structural residual field, the path segment covered by the abnormal residual cluster is determined as the positioning deviation region. The path feasible adjustment range of the defined current drilling point comprises, respectively expanding forward and backward with the graph structure node as the center, calculating the structural residual value and residual propagation weight between adjacent nodes; when the structural residual value of the adjacent node meets the decreasing trend, and the propagation weight of the corresponding edge is greater than the lower limit of the propagation weight, continuously selecting the graph structure node that meets the condition from the graph structure node to form the path feasible adjustment range.

6. A correction system based on the force feedback-based punch positioning strategy, based on the force feedback-based punch positioning strategy correction method according to any one of claims 1 to 5, characterized in that: The path modeling module is configured to obtain a target drilling path and discretize the path into a plurality of candidate drilling points in combination with device parameters; The stress prediction module is configured to construct a graph structure prediction model and generate a theoretical stress value of the candidate drilling point; The feedback processing module is configured to collect real-time force feedback data in the drilling process and compare the data with the theoretical stress value to obtain a structural residual; The residual analysis module is configured to construct a structural residual field, identify a positioning deviation area, and limit a path feasible adjustment range; The path correction module is configured to perform path backtracking reasoning in the adjustment range with the target drilling position as a guide to generate a posture correction path; The posture control module is configured to adjust a drilling posture according to the posture correction path and use the adjusted feedback data to update subsequent prediction and path correction conditions to realize continuous correction of the positioning strategy. 7.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the correction method of the drilling positioning strategy based on force feedback according to any one of claims 1-5.

8. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the correction method of the drilling positioning strategy based on force feedback according to any one of claims 1-5.

Images