Track plate drilling parameter intelligent optimization method and system based on deep learning

Abstract

The invention provides a track board drilling parameter intelligent optimization method and system based on deep learning, and relates to the technical field of deep learning, and the method comprises the steps: recognizing a feature region in three-dimensional scanning data through a deep learning semantic segmentation network, building a local coordinate system, calculating hole site deviation, constructing a pose correction field, and achieving the intelligent correction of drilling process parameters. And an optimal drilling track is generated by using a state space search algorithm. The method can adapt to actual geometric errors of the track plate, the drilling precision and efficiency are improved, the drilling failure rate is reduced, and the manufacturing quality of the track plate is improved.

Description

Technical Field

[0001] This invention relates to the field of deep learning technology, and in particular to a method and system for intelligent optimization of drilling parameters for track slabs based on deep learning. Background Technology

[0002] As a crucial component of high-speed railways and urban rail transit systems, the machining accuracy of track slabs directly affects the stability of the track structure and the safety of train operation. Drilling is a critical step in the manufacturing process of track slabs, requiring assurance of hole position accuracy, processing efficiency, and consistent quality. Traditional track slab drilling primarily relies on manual measurement and fixed mold positioning, followed by operator-controlled drilling equipment to complete the machining. With the rapid development of rail transit construction, automated drilling equipment is increasingly being applied to track slab production lines, enabling mass production through pre-programmed machining processes.

[0003] In recent years, with the development of computer vision and robotics technologies, vision-based automated drilling systems have begun to be applied in the track slab manufacturing industry. These systems acquire surface information of the track slab using cameras or laser scanning equipment, combine this information with image processing algorithms to determine drilling locations, and then control the drilling actuators via a CNC system to complete the machining task. This method offers greater adaptability and automation compared to traditional processes.

[0004] Existing technologies cannot accurately identify and adapt to the actual differences in the condition of track slabs with varying shapes and complex surface conditions, resulting in insufficient drilling position accuracy. This is especially true when the track slab has manufacturing deviations or deformation, as the preset drilling parameters often fail to meet actual processing requirements. Secondly, traditional drilling path planning lacks intelligent optimization capabilities and cannot dynamically adjust drilling parameters according to the actual position and orientation of the track slab, leading to low processing efficiency and low equipment utilization. Summary of the Invention

[0005] This invention provides a method and system for intelligent optimization of track slab drilling parameters based on deep learning, which can solve the problems in the prior art.

[0006] A first aspect of this invention provides a deep learning-based intelligent optimization method for track slab drilling parameters, comprising: The three-dimensional scanning data of the track slab and the preset drilling process data are acquired. A deep learning semantic segmentation network is used to identify the feature regions of the three-dimensional scanning data. The reference surface region and hole position region of the track slab are determined according to the region probability distribution corresponding to each feature. A local coordinate system for the track plate is established based on the reference surface area. Geometric center calculation and normal vector estimation are performed on the hole location area under the local coordinate system to determine the spatial position data of each hole location. By comparing the spatial position data with the standard hole location data in the drilling process data, the position deviation vector is calculated and a pose correction field is constructed. Based on the pose correction field, the drilling path parameters in the drilling process data are corrected point by point to obtain a drilling task point sequence that adapts to the actual track slab state; Using the drilling task point sequence as state nodes, and the joint space motion range of the drilling equipment, tool attitude reachability, and collision safety constraints as state transition conditions, an optimal drilling trajectory sequence satisfying all constraints is generated through a state space search algorithm. The drilling equipment is controlled to perform track slab drilling operations according to the optimal drilling trajectory sequence.

[0007] A deep learning semantic segmentation network is used to identify feature regions in the 3D scan data. Based on the probability distribution of each feature, the reference surface region and hole location region of the track slab are determined, including: The 3D scanning data is voxelized to obtain voxelized representation data. The voxelized representation data is then input into a deep learning semantic segmentation network. Multi-level feature extraction is performed on the voxelized representation data through convolution operations. The multi-level features are then upsampled and fused. The features at different levels are aligned in the spatial dimension and their channels are concatenated to obtain fused features. For each spatial location of the fused feature, a classification operation is performed to calculate the regional probability distribution of the location belonging to the reference surface category, the hole location category, and the background category. The category with the largest regional probability distribution is selected as the semantic label for that location. Based on the semantic tags, voxels with the same category tags and spatial connectivity are analyzed to identify the spatial range of each connected domain. The connected domain corresponding to the reference surface category tag is determined as the reference surface region, and the connected domains corresponding to the hole position category tag are determined as the hole position region.

[0008] A local coordinate system for the track slab is established based on the reference surface region, and the geometric center and normal vector of the hole location region are calculated and estimated under the local coordinate system to determine the spatial position data of each hole location, including: Point cloud data of the reference surface region is extracted, and plane fitting operation is performed on the point cloud data to obtain the fitted plane equation of the reference surface; The normal vector of the reference plane is extracted based on the fitted plane equation and used as the first coordinate axis direction of the local coordinate system. The geometric center point is selected as the origin of the local coordinate system from the point cloud data of the reference plane region. In a plane perpendicular to the first coordinate axis, the length direction of the track slab is defined as the second coordinate axis direction. The third coordinate axis direction is obtained by performing a vector cross product operation between the first and second coordinate axis directions, thus establishing the local coordinate system. The point cloud data of each sub-region of the hole location region is transformed from the global coordinate system to the local coordinate system; the centroid of each transformed point cloud data of each sub-region of the hole location is calculated to obtain the geometric center coordinates of each sub-region of the hole location in the local coordinate system. Surface fitting is performed on the point cloud data of each hole location sub-region, and the surface gradient direction of each hole location sub-region is calculated. The surface gradient direction is then used as the normal vector of each hole location sub-region. The spatial location data is formed by combining the geometric center coordinates and the normal vector of each hole.

[0009] Based on the pose correction field, the drilling path parameters in the drilling process data are corrected point by point to obtain a drilling task point sequence adapted to the actual track slab state, including: The design coordinates and design tool posture of each theoretical drilling point are extracted from the drilling process data; A spatial interpolation function is established in the pose correction field, which characterizes the mapping relationship between the pose deviation at any spatial location and the coordinates of that location. Input the design coordinates of each theoretical borehole point into the spatial interpolation function to calculate the position correction vector and attitude correction matrix at each theoretical borehole point; Apply the corresponding position correction vector to the design coordinates of each theoretical drilling point, and obtain the target coordinates of each corrected drilling point on the actual track slab surface through coordinate transformation calculation; Apply the corresponding attitude correction matrix to the designed tool attitude at each theoretical drilling point, and obtain the target tool attitude at each corrected drilling point through rotation transformation operation; Based on the target coordinates and target tool posture of each corrected drilling point, calculate the homogeneous transformation matrix of the tool posture at each corrected drilling point; According to the machining sequence constraints in the drilling process data, the tool pose homogeneous transformation matrix of each corrected drilling point is organized in sequence to form the drilling task point sequence.

[0010] Using the drilling task point sequence as state nodes, and the joint space motion range of the drilling equipment, tool attitude reachability, and collision safety constraints as state transition conditions, an optimal drilling trajectory sequence satisfying all constraints is generated through a state space search algorithm, including: Obtain the joint angle solution set corresponding to each drilling task point in the drilling task point sequence; determine whether each joint angle in the joint angle solution set is located between the upper and lower limits of the corresponding joint angle, and filter out the reachable drilling task points whose joint angles satisfy the range constraints; Calculate the angle between the actual cutter axis direction of the end effector at each reachable drilling task point and the target cutter axis direction required by the drilling task point, and filter out the executable drilling task points whose angle is less than a preset attitude deviation threshold; Obtain the geometric boundaries of the moving parts of the drilling equipment and the spatial envelope of obstacles in the working environment. For adjacent point pairs between the executable drilling task points, discretize and interpolate the joint angles from the starting point to the target point to generate several intermediate joint states. Calculate the spatial position of the geometric boundary of the moving part in each intermediate joint state, determine whether the spatial position overlaps with the spatial envelope of the obstacle, and determine whether there is a collision between the motion paths of adjacent point pairs; The motion path that does not involve collision is taken as the feasible motion path. The change in joint angle corresponding to each feasible motion path is calculated. The weighted sum of the changes in joint angle is taken as the motion cost of the feasible motion path. Starting from the initial drilling task point, a state-space search algorithm is used to search for path combinations that sequentially pass through all executable drilling task points in a collision-free connection relationship. The path combination with the minimum total motion cost of all feasible motion paths is selected as the optimal drilling trajectory sequence.

[0011] Starting from the initial drilling task point, a state-space search algorithm is used to search for path combinations that sequentially pass through all executable drilling task points in a collision-free connection relationship. The path combination with the minimum total motion cost of all feasible paths is selected as the optimal drilling trajectory sequence, including: Create a priority queue, wherein each element in the priority queue contains the identifier of the currently arrived borehole task point, the sequence of visited borehole task points, and the corresponding total movement cost; Pop the element with the lowest total movement cost from the priority queue as the current expansion element; Determine whether the access sequence of the current extended element already contains all executable drilling task points. If it does, mark the current extended element as a candidate solution and skip subsequent extended steps. If it does not, execute subsequent extended steps. Retrieve adjacent borehole task points that are directly connected to the borehole task point identifier of the current extended element and do not appear in the access sequence of the current extended element from the collision-free connection relationships; For each adjacent borehole task point, obtain the motion cost of a feasible motion path from the borehole task point of the current expanded element to the adjacent borehole task point, and add the motion cost to the total motion cost of the current expanded element to obtain the total motion cost after expansion; Repeat the step of popping the head element from the priority queue until the popped current extended element is marked as a candidate solution; The optimal candidate solution with the minimum total motion cost after expansion is selected from all candidate solutions, and the corresponding path combination is taken as the optimal borehole trajectory sequence.

[0012] A second aspect of the present invention provides a deep learning-based intelligent optimization system for track slab drilling parameters, comprising: The first unit is used to acquire three-dimensional scanning data of the track slab and preset drilling process data. It uses a deep learning semantic segmentation network to identify feature regions in the three-dimensional scanning data and determines the reference surface region and hole location region of the track slab based on the region probability distribution corresponding to each feature. The second unit is used to establish a local coordinate system for the track plate based on the reference surface area, and to perform geometric center calculation and normal vector estimation for the hole position area under the local coordinate system to determine the spatial position data of each hole position. By comparing the spatial position data with the standard hole position data in the drilling process data, the unit calculates the position deviation vector and constructs the pose correction field. The third unit is used to correct the drilling path parameters in the drilling process data point by point based on the pose correction field, so as to obtain a drilling task point sequence that adapts to the actual track slab state. The fourth unit is used to take the drilling task point sequence as state nodes, and the joint space motion range of the drilling equipment, tool posture reachability, and collision safety constraints as state transition conditions, and generate the optimal drilling trajectory sequence that satisfies all constraints through a state space search algorithm; The fifth unit is used to control the drilling equipment to perform track slab drilling operations according to the optimal drilling trajectory sequence.

[0013] A third aspect of the embodiments of the present invention, An electronic device is provided, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0014] Fourth aspect of the present invention, A computer-readable storage medium is provided, having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0015] The beneficial effects of this application are as follows: By using a deep learning semantic segmentation network to identify feature regions in 3D scan data, the reference surface area and hole location area of ​​the track slab can be automatically and accurately identified, overcoming the subjectivity and inefficiency of traditional manual identification methods and improving the accuracy of borehole location identification.

[0016] A local coordinate system for the track slab is established based on the identified reference surface region. On this basis, the geometric center of the hole position region is calculated and the normal vector is estimated, which realizes the accurate acquisition of spatial position data of each hole position and effectively copes with the deformation, displacement and other state deviations of the track slab in the actual environment.

[0017] By constructing a pose correction field and correcting the path parameters in the drilling process data point by point, the drilling parameters can adapt to the actual state of the track plate, ensuring drilling accuracy and reducing drilling errors caused by workpiece deviation.

[0018] By using the drilling task point sequence as state nodes, and comprehensively considering the joint space motion range of the drilling equipment, tool posture accessibility, and collision safety constraints, the optimal drilling trajectory sequence is generated through a state space search algorithm, thus solving the problem of drilling trajectory planning under complex constraints. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the intelligent optimization method for track slab drilling parameters based on deep learning, as described in an embodiment of the present invention. Figure 2 A flowchart illustrating the process of generating the optimal borehole trajectory sequence. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0022] refer to Figure 1 and Figure 2 The present invention provides a method for intelligent optimization of track slab drilling parameters based on deep learning, comprising: The three-dimensional scanning data of the track slab and the preset drilling process data are acquired. A deep learning semantic segmentation network is used to identify the feature regions of the three-dimensional scanning data. The reference surface region and hole position region of the track slab are determined according to the region probability distribution corresponding to each feature. A local coordinate system for the track plate is established based on the reference surface area. Geometric center calculation and normal vector estimation are performed on the hole location area under the local coordinate system to determine the spatial position data of each hole location. By comparing the spatial position data with the standard hole location data in the drilling process data, the position deviation vector is calculated and a pose correction field is constructed. Based on the pose correction field, the drilling path parameters in the drilling process data are corrected point by point to obtain a drilling task point sequence that adapts to the actual track slab state; Using the drilling task point sequence as state nodes, and the joint space motion range of the drilling equipment, tool attitude reachability, and collision safety constraints as state transition conditions, an optimal drilling trajectory sequence satisfying all constraints is generated through a state space search algorithm. The drilling equipment is controlled to perform track slab drilling operations according to the optimal drilling trajectory sequence.

[0023] In one optional implementation, a deep learning semantic segmentation network is used to identify feature regions in the 3D scan data, and the reference surface region and hole location region of the track slab are determined according to the region probability distribution corresponding to each feature, including: The 3D scanning data is voxelized to obtain voxelized representation data. The voxelized representation data is then input into a deep learning semantic segmentation network. Multi-level feature extraction is performed on the voxelized representation data through convolution operations. The multi-level features are then upsampled and fused. The features at different levels are aligned in the spatial dimension and their channels are concatenated to obtain fused features. For each spatial location of the fused feature, a classification operation is performed to calculate the regional probability distribution of the location belonging to the reference surface category, the hole location category, and the background category. The category with the largest regional probability distribution is selected as the semantic label for that location. Based on the semantic tags, voxels with the same category tags and spatial connectivity are analyzed to identify the spatial range of each connected domain. The connected domain corresponding to the reference surface category tag is determined as the reference surface region, and the connected domains corresponding to the hole position category tag are determined as the hole position region.

[0024] In this embodiment, a deep learning semantic segmentation network is used to identify the feature regions of the three-dimensional scanning data of the track slab. This mainly includes steps such as voxelization of the three-dimensional scanning data, feature extraction by the deep semantic segmentation network, feature fusion, classification operation, and connected component analysis, so as to accurately identify the reference surface region and hole region of the track slab.

[0025] The process involves acquiring 3D scan data of the track slab, typically obtained using a 3D laser scanner or structured light scanner. This data contains the track slab's geometric shape information and is represented by a large amount of 3D point cloud data. Because the raw 3D point cloud data is irregular in form, it is difficult to directly apply deep learning networks for processing; therefore, voxelization is required.

[0026] Voxelization is the process of converting irregular 3D point cloud data into a regular 3D mesh structure. In practice, the size of the voxels is first determined, for example, set to 2mm × 2mm × 2mm. Then, a uniform mesh is created in 3D space, and each point cloud data is assigned to a corresponding voxel based on its spatial location. If a point exists within a voxel, the voxel is marked as occupied; if no point exists, it is marked as idle. For voxels containing multiple points, the average position or other statistical characteristics of these points can be calculated as the attribute value of the voxel. After voxelization, the resulting voxel representation data is a 3D tensor, where each element represents the voxel state and attribute at the corresponding spatial location.

[0027] Voxelized representation data is input into a pre-trained deep learning semantic segmentation network. This network employs a three-dimensional convolutional neural network structure, comprising an encoder and a decoder. The encoder extracts features from the input data layer by layer through multiple layers of three-dimensional convolutional operations. Each convolutional layer is followed by batch normalization and a ReLU activation function to enhance the network's non-linear expressive power. The convolutional kernel size is typically set to 3×3×3, with a stride of 1 and padding of 1 to maintain the spatial dimension of the feature map. To reduce computational complexity and extract higher-level features, max pooling is applied after every few convolutional layers to halve the spatial resolution of the feature map.

[0028] Specifically, the encoder comprises multiple stages, each containing 2-3 convolutional layers. The number of channels in the output feature maps of each stage are 64, 128, 256, and 512, respectively, forming a hierarchical feature representation from low to high level. Low-level features mainly contain local information such as edges and textures, while high-level features contain more abstract semantic information. These features at different levels are crucial for accurately identifying the reference surface and hole locations, thus requiring feature fusion in the decoder.

[0029] The decoder restores the low-resolution feature maps to their original resolution through upsampling and then fuses them with the feature maps of the corresponding layers in the encoder. Upsampling is achieved using deconvolution (transposed convolution) with a kernel size of 2×2×2 and a stride of 2, doubling the spatial dimension of the feature maps. Feature fusion employs skip connections, concatenating the feature maps of the corresponding layers in the encoder with the upsampled feature maps along the channel dimension, combining low-level spatial details with high-level semantic information.

[0030] In the feature fusion process, feature maps from different levels are first upsampled to the same spatial resolution, and then concatenated along the channel dimension. For example, the feature map of the fourth stage (8×8×8×512) is upsampled to 16×16×16×512 and concatenated with the feature map of the third stage (16×16×16×256) to obtain a fused feature of 16×16×16×768. This process continues, upsampling layer by layer and fusing with the corresponding encoder layer features, ultimately obtaining a fused feature with the same spatial resolution as the input data, with the number of channels equal to the sum of the original feature channels.

[0031] For each spatial location of the fused features, a classification operation is performed, using a 1×1×1 convolutional kernel to map the fused features to class probabilities. Assuming the track slab recognition task includes three classes: reference surface class, hole location class, and background class, the convolutional kernel outputs 3 channels. The Softmax function is applied to normalize the three channel values ​​at each location, obtaining the probability distribution of that location belonging to each class. Based on the probability distribution, the class with the highest probability is selected as the semantic label for that location, forming a complete semantic segmentation result.

[0032] Connectivity analysis is performed on voxels with the same category label and spatial connectivity to identify the spatial extent of each connected region. A three-dimensional connected region labeling algorithm, such as a three-dimensional version of breadth-first search or region growing, is used to sequentially visit each voxel. For unvisited voxels labeled with the target category, neighboring voxels are recursively or iteratively visited. If neighboring voxels have the same category label, they are added to the current connected region until no further expansion is possible. In this way, all spatially connected voxels belonging to the same category can be grouped into a single connected region. The connected regions corresponding to the reference surface category label are identified as the reference surface region of the track slab; the connected regions corresponding to the hole position category labels are identified as the hole position regions of the track slab.

[0033] Through the above steps, the feature area identification of the three-dimensional scanning data of the track slab was completed, and the reference surface area and hole area were accurately located, providing an important spatial positioning basis for the subsequent track slab processing.

[0034] In one optional implementation, a local coordinate system for the track slab is established based on the reference surface region, and the geometric center and normal vector of the hole location region are calculated and estimated under the local coordinate system to determine the spatial position data of each hole location, including: Point cloud data of the reference surface region is extracted, and plane fitting operation is performed on the point cloud data to obtain the fitted plane equation of the reference surface; The normal vector of the reference plane is extracted based on the fitted plane equation and used as the first coordinate axis direction of the local coordinate system. The geometric center point is selected as the origin of the local coordinate system from the point cloud data of the reference plane region. In a plane perpendicular to the first coordinate axis, the length direction of the track slab is defined as the second coordinate axis direction. The third coordinate axis direction is obtained by performing a vector cross product operation between the first and second coordinate axis directions, thus establishing the local coordinate system. The point cloud data of each sub-region of the hole location region is transformed from the global coordinate system to the local coordinate system; the centroid of each transformed point cloud data of each sub-region of the hole location is calculated to obtain the geometric center coordinates of each sub-region of the hole location in the local coordinate system. Surface fitting is performed on the point cloud data of each hole location sub-region, and the surface gradient direction of each hole location sub-region is calculated. The surface gradient direction is then used as the normal vector of each hole location sub-region. The spatial location data is formed by combining the geometric center coordinates and the normal vector of each hole.

[0035] When determining the spatial location data of holes on the track plate, it is first necessary to establish a local coordinate system to accurately locate each hole. The reference surface area on the track plate serves as the basis for establishing the local coordinate system, while the hole location area is the target area whose spatial position needs to be determined.

[0036] Extracting point cloud data from the reference surface region is the first step in establishing a local coordinate system. A 3D laser scanning device is used to perform a full-range scan of the track slab, obtaining a complete point cloud dataset including the reference surface region. To improve subsequent processing efficiency, the original point cloud data can be downsampled using a voxel mesh filtering method. An appropriate voxel size (e.g., 0.5mm × 0.5mm × 0.5mm) is set, retaining one representative point within each voxel, thereby reducing the data volume while preserving geometric features.

[0037] When performing plane fitting on the point cloud data of the reference plane region, the RANSAC (Random Sample Consensus) algorithm is used for fitting calculation. This algorithm iteratively selects three points from the point cloud data to determine a plane, calculates the distance from the remaining points to the plane, and if the distance is less than a preset threshold (e.g., 0.2 mm), the point is considered an interior point. After multiple iterations (e.g., 1000 times), the plane with the most interior points is selected as the best fitting result, resulting in the plane equation Ax + By + Cz = 0, where the vector (A, B, C) is the plane normal vector.

[0038] Based on the fitted plane equation, the normal vector N=(A,B,C) is extracted as the direction of the first coordinate axis Z-axis of the local coordinate system. To ensure consistency of direction, the positive direction of the Z-axis is defined to face outward from the track plate. The geometric center of the point cloud in the reference plane region is used as the origin O of the coordinate system. The specific calculation method is to take the average of the coordinates of all points in the reference plane region: O=(∑xi / n,∑yi / n,∑zi / n), where n is the total number of points.

[0039] When determining the direction of the second coordinate axis, the actual installation orientation of the track slab is used, and the length direction of the track slab is taken as the X-axis direction. Principal Component Analysis (PCA) can be used to project the point cloud of the reference surface region onto a plane perpendicular to the Z-axis, and the direction of maximum change in the projected point set is calculated as the X-axis direction. After obtaining the X-axis direction vector X, the direction of the third coordinate axis Y-axis is obtained through the vector cross product operation Y=Z×X, thus establishing a complete local coordinate system.

[0040] Transforming the point cloud data of the borehole area from the global coordinate system to the local coordinate system is a crucial step in subsequent processing. Let the coordinates of the points in the global coordinate system be Pg=(xg,yg,zg), and the coordinates in the local coordinate system be Pl=(xl,yl,zl). The transformation relationship is: Pl=R·(Pg-O), where R is the rotation matrix composed of the three axis direction vectors of the local coordinate system, and O is the coordinate of the origin of the local coordinate system in the global coordinate system.

[0041] For each sub-region of the borehole location, point cloud data processing is required. First, Euclidean clustering is used to segment the transformed point cloud data, separating the point clouds of each sub-region of the borehole location. Then, the centroid coordinates of the point cloud of each sub-region of the borehole location are calculated as the geometric center of the borehole location: C=(∑xi / m,∑yi / m,∑zi / m), where m is the number of points in the point cloud of that sub-region of the borehole location.

[0042] Surface fitting is performed on each sub-region of the hole location to determine the normal vector. A quadratic surface fitting method is used, representing the hole location surface as z = ax² + by² + cxy + dx + ey + f. The parameters a, b, c, d, and e are solved using the least squares method to obtain the fitted surface. The gradient direction at any point (x, y, z) on the surface can be expressed as: After normalization, this becomes the normal vector direction at that point. The gradient direction at the geometric center of the hole is taken as the normal vector N for the entire hole.

[0043] Finally, the geometric center coordinates C and normal vector N of each hole are combined to form complete spatial location data. For holes with larger diameters, their effective diameter can also be calculated by extracting multiple points at the edge of the hole, fitting a circle or ellipse to obtain diameter information, further enhancing the completeness of the spatial location data.

[0044] In practical applications, to improve computational accuracy, point cloud data can be preprocessed, such as removing outliers and noise points. Outliers can be removed using statistical outlier filtering methods. For each point, the average distance to its nearest neighbors is calculated; if this distance exceeds three times the standard deviation of the global average distance, it is considered an outlier and removed. Furthermore, smoothing methods such as bilateral filtering can be used to reduce the impact of noise on the computational results while maintaining the clarity of edge features.

[0045] In one optional implementation, the drilling path parameters in the drilling process data are corrected point by point based on the pose correction field to obtain a drilling task point sequence adapted to the actual track slab state, including: The design coordinates and design tool posture of each theoretical drilling point are extracted from the drilling process data; A spatial interpolation function is established in the pose correction field, which characterizes the mapping relationship between the pose deviation at any spatial location and the coordinates of that location. Input the design coordinates of each theoretical borehole point into the spatial interpolation function to calculate the position correction vector and attitude correction matrix at each theoretical borehole point; Apply the corresponding position correction vector to the design coordinates of each theoretical drilling point, and obtain the target coordinates of each corrected drilling point on the actual track slab surface through coordinate transformation calculation; Apply the corresponding attitude correction matrix to the designed tool attitude at each theoretical drilling point, and obtain the target tool attitude at each corrected drilling point through rotation transformation operation; Based on the target coordinates and target tool posture of each corrected drilling point, calculate the homogeneous transformation matrix of the tool posture at each corrected drilling point; According to the machining sequence constraints in the drilling process data, the tool pose homogeneous transformation matrix of each corrected drilling point is organized in sequence to form the drilling task point sequence.

[0046] The design coordinates and tool orientation of each theoretical drilling point are extracted from the drilling process data. Drilling process data is typically stored in a database or standard format file, containing the theoretical coordinates (x, y, z) of each drilling point on the track slab, as well as tool orientation information. Tool orientation can be represented by a unit vector (hx, hy, hz) to indicate the drilling direction, or by Euler angles (α, β, γ) to indicate the tool's spatial orientation. The extraction process can be completed using a data parsing program, converting the structured data into a workable array format for easier subsequent processing.

[0047] A spatial interpolation function is established in the pose correction field. This function characterizes the mapping relationship between the pose deviation at any spatial location and the coordinates of that location. The pose correction field is constructed through the aforementioned calibration point measurements and includes the position deviation vector and attitude deviation matrix of each point in space. To achieve correction calculations for arbitrary locations, a three-dimensional spatial interpolation method, such as radial basis function (RBF) interpolation, is used. For position correction, the function F(x,y,z)=(Δx,Δy,Δz) is established; for attitude correction, the function G(x,y,z)=L is established, where L is the rotation correction matrix. The establishment of the interpolation function considers the spatial distribution characteristics of the calibration points, ensuring a smooth correction transition effect in the region between calibration points.

[0048] The design coordinates of each theoretical borehole point are input into a spatial interpolation function to calculate the position correction vector and attitude correction matrix at each theoretical borehole point. For each theoretical borehole point P_i(x_i,y_i,z_i), the position correction vector of that point is calculated using the position correction interpolation function F. ΔP_i=F(x_i,y_i,z_i)=(Δx_i,Δy_i,Δz_i); The attitude correction matrix L_i=G(x_i,y_i,z_i) for this point is calculated using the attitude correction interpolation function G. This step generates personalized correction parameters for each theoretical drilling point to adapt to the deformation state of the actual track slab.

[0049] Apply corresponding position correction vectors to the design coordinates of each theoretical drilling point, and obtain the target coordinates of each corrected drilling point on the actual track slab surface through coordinate transformation calculations. The formula for calculating the corrected drilling point coordinates P_i' is as follows: P_i'=P_i+ΔP_i, that is, (x_i',y_i',z_i')=(x_i+Δx_i,y_i+Δy_i,z_i+Δz_i).

[0050] This transformation maps the drilling points in the theoretical design space to the corresponding positions on the actual track slab surface, ensuring the precise positioning of the drilling locations.

[0051] A corresponding attitude correction matrix is ​​applied to the designed tool attitude at each theoretical drilling point, and the target tool attitude at each corrected drilling point is obtained through rotation transformation. Assuming the designed tool attitude at the theoretical drilling point is represented by a unit direction vector h_i, the corrected tool attitude h_i' is calculated using matrix multiplication: h_i' = L_i·h_i. If Euler angles are used to represent the attitude, they must first be converted to a rotation matrix, and after applying the attitude correction, the result is converted back to Euler angles. This step ensures that the tool entry angle remains perpendicular to the actual track plate surface or meets specific process angle requirements.

[0052] Based on the target coordinates and target tool posture of each corrected drilling point, the homogeneous transformation matrix of the tool posture at each corrected drilling point is calculated. The homogeneous transformation matrix provides a unified mathematical representation, facilitating subsequent posture control by the robot or CNC system.

[0053] Based on the machining sequence constraints in the drilling process data, the homogeneous transformation matrices of the tool pose for each corrected drilling point are organized sequentially to form a drilling task point sequence. The machining sequence constraints may consider various factors, such as minimizing the tool travel path, avoiding interference, and ensuring machining quality. The final drilling task point sequence can be represented as {T_1, T_2, ..., T_n}, where the subscripts indicate the execution order. This sequence is directly fed into the robot control system or CNC system to guide the drilling equipment to complete all drilling tasks along the optimized path.

[0054] In practical applications, to improve system efficiency, path smoothing processing can be added after the task point sequence is generated to ensure that the drilling equipment's movement trajectory between task points is smooth and continuous, avoiding drastic acceleration changes. Simultaneously, specific approach strategies, exit strategies, and drilling parameter settings can be added to each drilling point according to drilling process requirements, further improving drilling quality and efficiency.

[0055] Through the above detailed steps, the conversion process from the theoretical drilling path to the drilling task point sequence adapted to the actual track slab condition was completed, ensuring the accuracy and reliability of the drilling operation.

[0056] In one optional implementation, the drilling task point sequence is used as a state node, and the joint space motion range of the drilling equipment, tool posture reachability, and collision safety constraints are used as state transition conditions. An optimal drilling trajectory sequence satisfying all constraints is generated using a state space search algorithm, including: Obtain the joint angle solution set corresponding to each drilling task point in the drilling task point sequence; determine whether each joint angle in the joint angle solution set is located between the upper and lower limits of the corresponding joint angle, and filter out the reachable drilling task points whose joint angles satisfy the range constraints; Calculate the angle between the actual cutter axis direction of the end effector at each reachable drilling task point and the target cutter axis direction required by the drilling task point, and filter out the executable drilling task points whose angle is less than a preset attitude deviation threshold; Obtain the geometric boundaries of the moving parts of the drilling equipment and the spatial envelope of obstacles in the working environment. For adjacent point pairs between the executable drilling task points, discretize and interpolate the joint angles from the starting point to the target point to generate several intermediate joint states. Calculate the spatial position of the geometric boundary of the moving part in each intermediate joint state, determine whether the spatial position overlaps with the spatial envelope of the obstacle, and determine whether there is a collision between the motion paths of adjacent point pairs; The motion path that does not involve collision is taken as the feasible motion path. The change in joint angle corresponding to each feasible motion path is calculated. The weighted sum of the changes in joint angle is taken as the motion cost of the feasible motion path. Starting from the initial drilling task point, a state-space search algorithm is used to search for path combinations that sequentially pass through all executable drilling task points in a collision-free connection relationship. The path combination with the minimum total motion cost of all feasible motion paths is selected as the optimal drilling trajectory sequence.

[0057] In the operation of multi-axis drilling equipment, in order to improve drilling efficiency and ensure drilling quality, it is necessary to generate an optimal drilling trajectory sequence that meets various constraints based on the drilling task point sequence.

[0058] Obtain the joint angle solution set corresponding to each drilling task point in the drilling task point sequence. For each drilling task point, obtain multiple sets of possible joint angle solutions through inverse kinematics calculation. Inverse kinematics calculation can be performed using analytical methods or iterative methods. Analytical methods are suitable for robots with specific structures, while iterative methods, such as the Jacobian matrix rule, are suitable for complex structures. Taking a six-axis drilling robot as an example, each drilling task point may correspond to multiple sets of joint angle solutions. For example, the j-th drilling task point may have v solutions, denoted as {θj1, θj2, ..., θjv}, where θjv represents the v-th joint angle solution vector of the j-th task point, containing the angle values ​​of all joints.

[0059] Next, we check whether each joint angle in the joint angle solution set lies between the upper and lower limits of the corresponding joint angle, and filter out reachable drilling task points whose joint angles satisfy the range constraints. For each joint angle solution vector θjv, we check whether each joint angle θjvk satisfies the constraint: θmin,k ≤ θjvk ≤ θmax,k, where θmin,k and θmax,k represent the lower and upper limits of the k-th joint angle, respectively. Only when all joint angles satisfy the constraints is the solution considered valid, and the corresponding drilling task point considered reachable.

[0060] Calculate the angle between the actual tool axis direction of the end effector at each reachable drilling task point and the target tool axis direction required for that drilling task point, and select executable drilling task points with an angle smaller than a preset attitude deviation threshold. Assuming the target tool axis direction is a unit vector vt, and the actual tool axis direction corresponding to the joint angle solution θjv is a unit vector va, the angle α can be calculated using the formula α = arccos(vt·va). Typically, the preset attitude deviation threshold is set to a small angle value, such as 0.5 degrees or 1 degree, to ensure drilling accuracy.

[0061] After determining the feasible drilling task points, the geometric boundaries of the moving parts of the drilling equipment and the spatial envelope of obstacles in the working environment are obtained. The moving parts of the drilling equipment can be modeled using a series of geometric shapes (such as cylinders, cuboids, etc.), while the spatial envelope of obstacles can be represented using point cloud data or predefined geometric models.

[0062] For adjacent point pairs between executable drilling task points, the joint angles from the starting point to the target point are discretized and interpolated to generate several intermediate joint states. The interpolation method typically uses linear interpolation, i.e., uniform sampling is performed between the initial joint state θs and the target joint state θg at a certain step size. The step size should be chosen to balance computational efficiency with collision detection accuracy; for example, it can be set to sample joint angle changes every 10 milliseconds or every 0.1 radians.

[0063] Next, the spatial position of the geometric boundaries of moving parts in each intermediate joint state is calculated. It is then determined whether this spatial position overlaps with the spatial envelope of obstacles, thus identifying potential collisions between adjacent point pairs. Specifically, for each intermediate joint state, the position and orientation of each link are calculated based on forward kinematics, and then these geometric boundaries are checked for intersections with obstacles in the environment. Collision detection algorithms can employ layered bounding box techniques or the GJK algorithm to improve detection efficiency.

[0064] The motion path that does not involve collision is considered a feasible motion path. The joint angle changes corresponding to each feasible motion path are calculated, and the weighted sum of the joint angle changes is taken as the motion cost of the feasible motion path. The motion cost can be expressed as: C = w1|Δθ1| + w2|Δθ2| + ... + wv|Δθv|, where Δθj represents the angle change of the j-th joint, and wj is the corresponding weight coefficient. The weight coefficient is usually related to the joint's moment of inertia, energy consumption, or importance. For example, the base joint usually has a larger weight, while the distal end joints have a smaller weight.

[0065] Finally, starting from the initial drilling task point, a state-space search algorithm is used to search for path combinations that sequentially pass through all executable drilling task points in a collision-free connection relationship. The path combination with the minimum total motion cost of all feasible paths is selected as the optimal drilling trajectory sequence. The search algorithm can employ... The algorithm or Dijkstra's algorithm, where The algorithm uses a heuristic function to guide the search direction, improving efficiency. During the search process, a priority queue can be used to store the paths to be explored, sorting them according to the current cost of the path plus the estimated cost to the target, prioritizing the exploration of paths with lower costs.

[0066] In practical applications, algorithm parameters can be adjusted for specific drilling tasks. For example, for drilling aerospace structural components with high precision requirements, the attitude deviation threshold can be set smaller; for batch drilling tasks that need to be completed quickly, the weight coefficients in the motion cost calculation can be appropriately adjusted to emphasize execution efficiency. Meanwhile, the choice of state-space search algorithm can also be flexibly adjusted according to the task scale. Exhaustive search can be used for small-scale tasks, while heuristic or optimization algorithms, such as genetic algorithms or simulated annealing algorithms, are more suitable for large-scale tasks to find an approximate optimal solution within a reasonable time.

[0067] In one optional implementation, starting from the initial drilling task point, a state-space search algorithm is used to search for path combinations that sequentially pass through all executable drilling task points in a collision-free connection relationship. The path combination with the minimum total motion cost of all feasible motion paths is selected as the optimal drilling trajectory sequence, including: Create a priority queue, wherein each element in the priority queue contains the identifier of the currently arrived borehole task point, the sequence of visited borehole task points, and the corresponding total movement cost; Pop the element with the lowest total movement cost from the priority queue as the current expansion element; Determine whether the access sequence of the current extended element already contains all executable drilling task points. If it does, mark the current extended element as a candidate solution and skip subsequent extended steps. If it does not, execute subsequent extended steps. Retrieve adjacent borehole task points that are directly connected to the borehole task point identifier of the current extended element and do not appear in the access sequence of the current extended element from the collision-free connection relationships; For each adjacent borehole task point, obtain the motion cost of a feasible motion path from the borehole task point of the current expanded element to the adjacent borehole task point, and add the motion cost to the total motion cost of the current expanded element to obtain the total motion cost after expansion; Repeat the step of popping the head element from the priority queue until the popped current extended element is marked as a candidate solution; The optimal candidate solution with the minimum total motion cost after expansion is selected from all candidate solutions, and the corresponding path combination is taken as the optimal borehole trajectory sequence.

[0068] The search process begins by creating a priority queue. Each element in this queue contains three key pieces of information: the identifier of the currently arrived borehole task point, the sequence of visited borehole task points, and the corresponding total movement cost. During queue initialization, the starting borehole task point is inserted as the first element. At this point, the visit sequence only contains the starting point, and the total movement cost is zero.

[0069] Subsequently, the element with the lowest total movement cost is popped from the priority queue and used as the current expansion element. The priority queue is sorted based on the total movement cost, ensuring that the path with the lowest cost is selected for expansion each time, which conforms to the principle of optimal search.

[0070] Next, determine whether the access sequence of the current extended element contains all executable drilling task points. If it contains all task points, it indicates that a complete path has been found, and the extended element is marked as a candidate solution and the subsequent extended steps are skipped; if it does not contain all task points, the extended operation continues.

[0071] For elements that need further expansion, adjacent drilling task points that are directly connected to the current drilling task point and do not appear in the current access sequence are retrieved from the pre-established collision-free connectivity relationships. The connectivity relationships can be pre-calculated using robotic arm kinematics and collision detection algorithms and stored as an adjacency matrix or adjacency list.

[0072] For each connectable adjacent borehole task point, obtain the motion cost of a feasible path from the current task point to that adjacent point. Motion cost can be measured by metrics such as path length, movement time, or energy consumption. Add this motion cost to the total motion cost of the current extended element to obtain the expanded total motion cost. Then, construct a new extended element containing the identifier of the newly arrived borehole task point, the updated access sequence (adding the new task point to the end of the sequence), and the updated total motion cost. Insert this new element into the priority queue.

[0073] Repeat the step of popping the head element from the priority queue until the popped current expansion element is marked as a candidate solution, indicating that a complete path containing all drilling task points has been found. Due to the characteristics of the priority queue, if multiple paths containing all task points exist, they will be found sequentially in ascending order of total movement cost.

[0074] Finally, the optimal candidate solution with the minimum total motion cost is selected from all candidate solutions, and its corresponding path combination is taken as the optimal drilling trajectory sequence. This sequence will guide the robotic arm to complete all drilling tasks in the optimal order.

[0075] In practical applications, suppose a sheet metal part requires 10 drilling operations, with multiple possible movement paths between each drilling point. Using the algorithm described above, the drilling sequence with the shortest total movement time can be quickly found, taking into account robotic arm motion constraints and environmental obstacles.

[0076] Priority queues can be implemented using binary heaps or Fibonacci heaps to improve queue operation efficiency. For each expansion step, if the current access sequence length is K and there are Y drilling task points, then at most (YK) adjacent points need to be checked. Appropriate pruning strategies, such as discarding an expansion element when its total movement cost exceeds the known optimal solution, can further improve algorithm efficiency.

[0077] When dealing with large-scale drilling tasks, a heuristic function can be introduced to estimate the cost from the current state to the target state, thus extending the basic search algorithm to... Searching can further improve search efficiency. Heuristic functions can be designed based on the minimum spanning tree cost or the nearest neighbor distance between remaining unvisited points.

[0078] By combining state-space search with robotic arm kinematics and path planning techniques, this method can effectively find the optimal execution sequence for multi-point drilling tasks in complex industrial environments, thereby improving production efficiency and reducing energy consumption.

[0079] This invention provides a deep learning-based intelligent optimization system for track slab drilling parameters, comprising: The first unit is used to acquire three-dimensional scanning data of the track slab and preset drilling process data. It uses a deep learning semantic segmentation network to identify feature regions in the three-dimensional scanning data and determines the reference surface region and hole location region of the track slab based on the region probability distribution corresponding to each feature. The second unit is used to establish a local coordinate system for the track plate based on the reference surface area, and to perform geometric center calculation and normal vector estimation for the hole position area under the local coordinate system to determine the spatial position data of each hole position. By comparing the spatial position data with the standard hole position data in the drilling process data, the unit calculates the position deviation vector and constructs the pose correction field. The third unit is used to correct the drilling path parameters in the drilling process data point by point based on the pose correction field, so as to obtain a drilling task point sequence that adapts to the actual track slab state. The fourth unit is used to take the drilling task point sequence as state nodes, and the joint space motion range of the drilling equipment, tool posture reachability, and collision safety constraints as state transition conditions, and generate the optimal drilling trajectory sequence that satisfies all constraints through a state space search algorithm; The fifth unit is used to control the drilling equipment to perform track slab drilling operations according to the optimal drilling trajectory sequence.

[0080] A third aspect of the present invention, An electronic device is provided, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0081] Fourth aspect of the present invention, A computer-readable storage medium is provided, having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0082] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent optimization of drilling parameters for track slabs based on deep learning, characterized in that, include: The three-dimensional scanning data of the track slab and the preset drilling process data are acquired. A deep learning semantic segmentation network is used to identify the feature regions of the three-dimensional scanning data. The reference surface region and hole position region of the track slab are determined according to the region probability distribution corresponding to each feature. A local coordinate system for the track plate is established based on the reference surface area. Geometric center calculation and normal vector estimation are performed on the hole location area under the local coordinate system to determine the spatial position data of each hole location. By comparing the spatial position data with the standard hole location data in the drilling process data, the position deviation vector is calculated and a pose correction field is constructed. Based on the pose correction field, the drilling path parameters in the drilling process data are corrected point by point to obtain a drilling task point sequence that adapts to the actual track slab state; Using the drilling task point sequence as state nodes, and the joint space motion range of the drilling equipment, tool attitude reachability, and collision safety constraints as state transition conditions, an optimal drilling trajectory sequence satisfying all constraints is generated through a state space search algorithm. The drilling equipment is controlled to perform track slab drilling operations according to the optimal drilling trajectory sequence.

2. The method according to claim 1, characterized in that, A deep learning semantic segmentation network is used to identify feature regions in the 3D scan data. Based on the probability distribution of each feature, the reference surface region and hole location region of the track slab are determined, including: The 3D scanning data is voxelized to obtain voxelized representation data. The voxelized representation data is then input into a deep learning semantic segmentation network. Multi-level feature extraction is performed on the voxelized representation data through convolution operations. The multi-level features are then upsampled and fused. The features at different levels are aligned in the spatial dimension and their channels are concatenated to obtain fused features. For each spatial location of the fused feature, a classification operation is performed to calculate the regional probability distribution of the location belonging to the reference surface category, the hole location category, and the background category. The category with the largest regional probability distribution is selected as the semantic label for that location. Based on the semantic tags, voxels with the same category tags and spatial connectivity are analyzed to identify the spatial range of each connected domain. The connected domain corresponding to the reference surface category tag is determined as the reference surface region, and the connected domains corresponding to the hole position category tag are determined as the hole position region.

3. The method according to claim 1, characterized in that, A local coordinate system for the track slab is established based on the reference surface region, and the geometric center and normal vector of the hole location region are calculated and estimated under the local coordinate system to determine the spatial position data of each hole location, including: Point cloud data of the reference surface region is extracted, and plane fitting operation is performed on the point cloud data to obtain the fitted plane equation of the reference surface; The normal vector of the reference plane is extracted based on the fitted plane equation and used as the first coordinate axis direction of the local coordinate system. The geometric center point is selected as the origin of the local coordinate system from the point cloud data of the reference plane region. In a plane perpendicular to the first coordinate axis, the length direction of the track slab is defined as the second coordinate axis direction. The third coordinate axis direction is obtained by performing a vector cross product operation between the first and second coordinate axis directions, thus establishing the local coordinate system. The point cloud data of each sub-region of the hole location region is transformed from the global coordinate system to the local coordinate system; the centroid of each transformed point cloud data of each sub-region of the hole location is calculated to obtain the geometric center coordinates of each sub-region of the hole location in the local coordinate system. Surface fitting is performed on the point cloud data of each hole location sub-region, and the surface gradient direction of each hole location sub-region is calculated. The surface gradient direction is then used as the normal vector of each hole location sub-region. The spatial location data is formed by combining the geometric center coordinates and the normal vector of each hole.

4. The method according to claim 1, characterized in that, Based on the pose correction field, the drilling path parameters in the drilling process data are corrected point by point to obtain a drilling task point sequence adapted to the actual track slab state, including: The design coordinates and design tool posture of each theoretical drilling point are extracted from the drilling process data; A spatial interpolation function is established in the pose correction field, which characterizes the mapping relationship between the pose deviation at any spatial location and the coordinates of that location. Input the design coordinates of each theoretical borehole point into the spatial interpolation function to calculate the position correction vector and attitude correction matrix at each theoretical borehole point; Apply the corresponding position correction vector to the design coordinates of each theoretical drilling point, and obtain the target coordinates of each corrected drilling point on the actual track slab surface through coordinate transformation calculation; Apply the corresponding attitude correction matrix to the designed tool attitude at each theoretical drilling point, and obtain the target tool attitude at each corrected drilling point through rotation transformation operation; Based on the target coordinates and target tool posture of each corrected drilling point, calculate the homogeneous transformation matrix of the tool posture at each corrected drilling point; According to the machining sequence constraints in the drilling process data, the tool pose homogeneous transformation matrix of each corrected drilling point is organized in sequence to form the drilling task point sequence.

5. The method according to claim 1, characterized in that, Using the drilling task point sequence as state nodes, and the joint space motion range of the drilling equipment, tool attitude reachability, and collision safety constraints as state transition conditions, an optimal drilling trajectory sequence satisfying all constraints is generated through a state space search algorithm, including: Obtain the joint angle solution set corresponding to each drilling task point in the drilling task point sequence; determine whether each joint angle in the joint angle solution set is located between the upper and lower limits of the corresponding joint angle, and filter out the reachable drilling task points whose joint angles satisfy the range constraints; Calculate the angle between the actual cutter axis direction of the end effector at each reachable drilling task point and the target cutter axis direction required by the drilling task point, and filter out the executable drilling task points whose angle is less than a preset attitude deviation threshold; Obtain the geometric boundaries of the moving parts of the drilling equipment and the spatial envelope of obstacles in the working environment. For adjacent point pairs between the executable drilling task points, discretize and interpolate the joint angles from the starting point to the target point to generate several intermediate joint states. Calculate the spatial position of the geometric boundary of the moving part in each intermediate joint state, determine whether the spatial position overlaps with the spatial envelope of the obstacle, and determine whether there is a collision between the motion paths of adjacent point pairs; The motion path that does not involve collision is taken as the feasible motion path. The change in joint angle corresponding to each feasible motion path is calculated. The weighted sum of the changes in joint angle is taken as the motion cost of the feasible motion path. Starting from the initial drilling task point, a state-space search algorithm is used to search for path combinations that sequentially pass through all executable drilling task points in a collision-free connection relationship. The path combination with the minimum total motion cost of all feasible motion paths is selected as the optimal drilling trajectory sequence.

6. The method according to claim 5, characterized in that, Starting from the initial drilling task point, a state-space search algorithm is used to search for path combinations that sequentially pass through all executable drilling task points in a collision-free connection relationship. The path combination with the minimum total motion cost of all feasible motion paths is selected as the optimal drilling trajectory sequence, including: Create a priority queue, wherein each element in the priority queue contains the identifier of the currently arrived borehole task point, the sequence of visited borehole task points, and the corresponding total movement cost; Pop the element with the lowest total movement cost from the priority queue as the current expansion element; Determine whether the access sequence of the current extended element already contains all executable drilling task points. If it does, mark the current extended element as a candidate solution and skip subsequent extended steps. If it does not, execute subsequent extended steps. Retrieve adjacent borehole task points that are directly connected to the borehole task point identifier of the current extended element and do not appear in the access sequence of the current extended element from the collision-free connection relationships; For each adjacent borehole task point, obtain the motion cost of a feasible motion path from the borehole task point of the current expanded element to the adjacent borehole task point, and add the motion cost to the total motion cost of the current expanded element to obtain the total motion cost after expansion; Repeat the step of popping the head element from the priority queue until the popped current extended element is marked as a candidate solution; The optimal candidate solution with the minimum total motion cost after expansion is selected from all candidate solutions, and the corresponding path combination is taken as the optimal borehole trajectory sequence.

7. A deep learning-based intelligent optimization system for track slab drilling parameters, used to implement the method as described in any one of claims 1-6, characterized in that, include: The first unit is used to acquire three-dimensional scanning data of the track slab and preset drilling process data. It uses a deep learning semantic segmentation network to identify feature regions in the three-dimensional scanning data and determines the reference surface region and hole location region of the track slab based on the region probability distribution corresponding to each feature. The second unit is used to establish a local coordinate system for the track plate based on the reference surface area, and to perform geometric center calculation and normal vector estimation for the hole position area under the local coordinate system to determine the spatial position data of each hole position. By comparing the spatial position data with the standard hole position data in the drilling process data, the unit calculates the position deviation vector and constructs the pose correction field. The third unit is used to correct the drilling path parameters in the drilling process data point by point based on the pose correction field, so as to obtain a drilling task point sequence that adapts to the actual track slab state. The fourth unit is used to take the drilling task point sequence as state nodes, and the joint space motion range of the drilling equipment, tool posture reachability, and collision safety constraints as state transition conditions, and generate the optimal drilling trajectory sequence that satisfies all constraints through a state space search algorithm; The fifth unit is used to control the drilling equipment to perform track slab drilling operations according to the optimal drilling trajectory sequence.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images