Glass deep hole laser drilling track planning method and system based on ant colony algorithm

Abstract

The invention discloses a glass deep hole laser drilling track planning method and system based on an ant colony algorithm, and the method comprises the following steps: obtaining a to-be-processed task parameter, a kinematics constraint parameter and a process constraint parameter, and taking the parameters as an input parameter set; a track template set and a process parameter candidate set are established, a data acquisition channel is monitored online, and a trigger threshold value is set; generating a candidate decision set based on the to-be-processed task parameters, the depth segmentation, the track template set and the process parameter candidate set; establishing a comprehensive cost evaluation system; an improved Beam-ACO algorithm is adopted on the candidate decision set, and a partial solution set with the optimal score is reserved; executing hole sequence selection and segmented trajectory selection in the partial solution set to obtain a hole sequence planning result and a segmented trajectory planning result; and generating an executable scanning track and a platform motion path according to the hole sequence planning result, the segmented track planning result and the input parameter set. According to the invention, the glass deep hole laser drilling processing efficiency and the track feasibility are improved.

Description

Technical Field

[0001] This invention relates to the field of laser precision machining technology, and in particular to a method and system for planning laser drilling trajectories for deep holes in glass based on ant colony algorithm. Background Technology

[0002] Glass materials, due to their high hardness and brittleness, low thermal conductivity, and sensitivity to microcracks, are prone to problems such as edge chipping, microcrack propagation, fluctuations in hole wall roughness, and slag shielding caused by poor slag removal during micro-hole array and deep hole laser drilling. Existing glass laser drilling processes typically require high consistency in processing under conditions of a large number of holes, large hole depths, and small hole diameters. This not only places demands on the feasibility of laser scanning trajectories and platform movement paths but also requires balancing thermal impact and hole shape quality control under multiple constraints, including processing cycle time, dwell cooling, reversal and transition constraints, and focusing range.

[0003] In existing technologies, the common practice is for process engineers to set the hole processing sequence and single-hole scanning strategy based on experience, or to use simple geometric optimization methods such as shortest distance and traveling salesman path to shorten the platform's movement distance. For single-hole deep hole processing, a fixed circular or spiral scanning trajectory is usually selected, and power, speed, or focal position is adjusted according to preset rules during processing. Some solutions introduce intelligent optimization methods such as ant colony algorithms to optimize the hole sequence, but most of them only limit the optimization objective to travel or time, or separate the hole sequence optimization from the single-hole segmented trajectory and parameter optimization into two independent problems to be solved separately, lacking a unified candidate decision modeling and comprehensive cost evaluation system. At the same time, common optimization processes are mostly completed offline. If quality risks caused by occlusion, abnormal slag discharge, or heat accumulation occur during the execution phase, it is often necessary to manually stop the machine to adjust parameters or restart the task, making it difficult to transform online monitoring signals into a feedback mechanism that can be used for trajectory replanning.

[0004] Therefore, how to provide a method and system for planning laser drilling trajectories for deep holes in glass based on ant colony algorithms is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a method and system for planning the trajectory of deep-hole laser drilling in glass based on the ant colony algorithm. This invention constructs a joint decision model of hole sequence and depth segmented trajectory, uses an improved Beam-ACO to perform two-stage beam search on the candidate decision set and prunes to retain the optimal partial solution, and combines online monitoring-triggered adaptive update of comprehensive cost to achieve rapid output of executable scanning trajectory and platform path, thereby improving the processing efficiency and trajectory feasibility of deep-hole laser drilling in glass, and enhancing slag removal stability and hole shape consistency.

[0006] The method and system for planning laser drilling trajectory for deep holes in glass based on ant colony algorithm according to embodiments of the present invention include the following steps: Obtain the task parameters, kinematic constraint parameters, and process constraint parameters of the glass workpiece as the input parameter set; Each hole to be processed is segmented into depths according to the parameters of the task to be processed, and a set of trajectory templates, a set of candidate process parameters, an online monitoring data acquisition channel and a trigger threshold are set. A candidate decision set is generated based on the parameters of the task to be processed, depth segmentation, trajectory template set, and candidate set of process parameters. Based on the candidate decision set, initialize the pheromone distribution and heuristic information, and establish a comprehensive cost evaluation system; An improved Beam-ACO algorithm is used on the candidate decision set to perform a two-level beam search of aperture order beam search and depth segmented trajectory beam search. At each decision level, the partial solutions formed by aperture order selection are expanded and pruned, and the set of partial solutions with the best scores is retained according to the preset beam width. In a partial solution set, hole sequence selection and segmented trajectory selection are performed. Online monitoring data is collected and threshold judgment is performed. When the judgment is valid, the comprehensive cost evaluation system is updated to obtain hole sequence planning results and segmented trajectory planning results. Based on the hole sequence planning results, segmented trajectory planning results, and input parameter set, an executable scanning trajectory and platform motion path are generated.

[0007] Optionally, obtaining the input parameter set specifically includes: The system receives the processing task parameters of the glass workpiece to be processed and writes them into the task data structure. The task data structure includes the number of holes, hole position coordinates, target hole diameter, minimum spacing between holes, and target hole depth. The number of holes and the number of hole position coordinate entries are checked to be consistent. The target hole diameter and target hole depth are checked one by one and are positive. The task data that passes the check is locked and saved. Receive the kinematic constraint parameters of the glass workpiece to be processed, establish an association index between the kinematic constraint parameters and the task data structure, and save it; Receive the process constraint parameters of the glass workpiece to be processed, and merge the process constraint parameters with the task data structure and kinematic constraint parameters to generate an input parameter set.

[0008] Optionally, the establishment of the trajectory template set, the candidate set of process parameters, the online monitoring data acquisition channel, and the setting of the trigger threshold specifically include: Based on the parameters of the task to be processed, the target hole depth is read one by one, the segmented depth interval is set, and boundary points are generated from zero depth according to the segmented depth interval. Adjacent boundary points are paired to generate depth segments. The depth segments are numbered according to the generation order and the start and end depths of each segment are recorded. The depth segmentation results are output. Based on the depth segmentation results, a trajectory template is generated sequentially for each depth segment. The template number and adjustable parameter bits are written into each template. The trajectory template is then bound to the corresponding depth segment and stored. A set of trajectory templates is output. Based on the depth segmentation results, candidate parameter values ​​are generated sequentially for each depth segment and combined to form a candidate parameter group. For each candidate parameter group, the input parameter set is called to perform verification and remove candidate parameter groups that fail the verification. An online monitoring data acquisition channel is constructed and a trigger threshold is configured. The trigger threshold is associated with and stored with the corresponding depth segment.

[0009] Optionally, the generation of the candidate decision set specifically includes: Assign a hole number to each hole to be processed, traverse any two holes and calculate the distance between them, compare the distance between holes with the minimum spacing between holes, and write the hole pairs that meet the spacing requirements into the hole sequence candidate table. Read the depth segmentation results hole by hole and segment by segment, read the trajectory template set and process parameter candidate set bound to each depth segment, pair each trajectory template and each candidate parameter group to generate depth segment candidate pairs, and write them into the depth segment candidate table. Traverse each hole sequence transfer candidate in the hole sequence candidate table, read the starting hole number and the target hole number, traverse the depth segment candidate table entries corresponding to the target hole and read the segment number, trajectory template identifier and candidate parameter group identifier, and combine the hole sequence transfer candidate with the segment number, trajectory template identifier and candidate parameter group identifier to obtain the candidate decision set.

[0010] Optionally, the comprehensive cost evaluation system is obtained by specifically including: Read the candidate decision set table and iterate through each one, assign a candidate decision number to each candidate decision, write it into the candidate index table, and output the candidate index table. The pheromone distribution is initialized based on the candidate index table, the pore order pheromone table and the segmented pheromone table are established, and a unified initial pheromone value is written for each pore order transition candidate and candidate decision. The pore order pheromone table and the segmented pheromone table are then output. For each candidate decision, the processing time cost and kinematic constraint penalty are calculated and written in sequence. After writing each cost component, the cost components are weighted and summed according to the preset weights to obtain the comprehensive cost value. The comprehensive cost value is then processed by taking the reciprocal and adding a preset positive number to obtain the heuristic information value, which is written into the heuristic information table. The comprehensive cost evaluation system is then output.

[0011] Optionally, obtaining the set of partial solutions with the optimal score specifically includes: An improved Beam-ACO algorithm is used on the candidate decision set. The candidate decision set, pheromone distribution and heuristic information are read, a partial solution data structure is established and empty hole sequence and empty segment trajectory are written. The current decision layer is set as the starting layer. The partial solution containing only empty hole sequence and empty segment trajectory is written into the partial solution set and output as the initial partial solution set. The initial partial solution set is copied to the ant construction queue corresponding to the number of ants and the current partial solution set of each ant is initialized. The hole order search is performed using a two-level beam search. The initial partial solution set of the current decision layer is traversed one by one in an ant-constructed queue. For each partial solution of each ant, the selected hole set and the current last hole number are read. Hole order candidates with the starting hole number equal to the current last hole number and the target hole number not appearing in the selected hole set are selected from the candidate decision set. For each hole order candidate, the corresponding pheromone value and the corresponding heuristic information value are read. The selection weight of the hole order candidate is calculated according to the pheromone value and the heuristic information value, and all selection weights are normalized to obtain the selection probability. Sampling is performed according to the selection probability or the target hole number is determined according to the maximum probability rule. The target hole number is appended to the partial solution to form the hole order extended partial solution. When forming the hole order extended partial solution, the transfer time is included in the processing time cost. Then the hole order extended partial solution is written into the extended set. The depth segmented trajectory bundle search is performed using a two-stage bundle search. The extended set is traversed one by one. For each hole sequence extension partial solution, the newly added target hole number and the depth segmentation result of the target hole are read. Segmented trajectory candidates that match the target hole number and depth segment number are selected from the candidate decision set according to the depth segment number. Feasibility verification is performed on each segmented trajectory candidate to eliminate candidates that violate kinematic constraint parameters and process constraint parameters. For candidates that fail the verification but can be repaired, repair operations are performed. For segmented trajectory candidates that pass the verification or are repaired, the incremental cost is calculated one by one and added to the determined cost of the partial solution to form a segmented extension score. Segmented trajectory candidates for each depth segment are selected from best to worst according to the segmented extension score and combined to form different segmented extension partial solutions. The segmented extension partial solutions are written into the score extension set. The score extension set is sorted from best to worst and pruned. The score is obtained by weighting the processing time cost and kinematic constraint penalty. An update operation is performed on the pheromone distribution. The pore order pheromone and segmented pheromone are proportionally attenuated according to the pheromone evaporation parameter. The pore order selection and segmented trajectory selection that appear in the partial solution with the best score are written with pheromone increment parameter. The set of partial solutions with the best score is output.

[0012] Optionally, obtaining the hole sequence planning result and the segmented trajectory planning result specifically includes: Read a partial solution set and use it as the current candidate set. Perform hole order selection and segmented trajectory selection on the current candidate set. At the same time, perform non-dominated sorting to extract non-dominated solutions. If the limit is exceeded, filter by crowding distance and write it into the segmented trajectory field. Output the hole order candidate set. The candidate set of hole order is determined with the corresponding trigger threshold. If any trigger threshold is exceeded, a trigger mark is generated. The comprehensive cost evaluation system is updated for the piecewise extended partial solution with the trigger mark, and the processing time cost and kinematic constraint penalty of the piecewise extended partial solution are updated simultaneously. The sorting and pruning are re-executed and the hole order field and piecewise trajectory field are read to obtain the hole order planning result and the piecewise trajectory planning result. Optionally, the generation of the executable scanning trajectory and platform motion path specifically includes: Read the hole sequence planning results, segmented trajectory planning results and input parameter set, write them together into the generated execution plan data and archive them; Read the hole sequence and hole position coordinates from the execution plan data, set the initial position of the platform as the path start point and write it into the path point sequence, write the hole position coordinates into the path point sequence as the platform target point, calculate the transfer distance between adjacent points in the path point sequence segment by segment, write the start point, end point, transfer distance and transfer time of each transfer segment into the platform motion segment record in sequence, and output the platform path and scheduling table. Read and verify the process constraint parameters and kinematic constraint parameters one hole at a time and segment at a time, generate the scan trajectory point sequence and encapsulation sequence according to the template, and export the scan trajectory file and platform path file.

[0013] A glass deep hole laser drilling trajectory planning system based on ant colony algorithm according to an embodiment of the present invention includes: The parameter acquisition module is used to acquire the processing parameters, kinematic constraint parameters, and process constraint parameters of the glass workpiece. The segmented database module is used to establish a set of trajectory templates and a set of candidate process parameters, as well as to establish an online monitoring data acquisition channel and set trigger thresholds. The candidate generation module is used to generate a candidate decision set based on the parameters of the task to be processed, depth segmentation, trajectory template set, and process parameter candidate set. The evaluation initialization module is used to initialize the pheromone distribution and heuristic information based on the candidate decision set, and to establish a comprehensive cost evaluation system. The two-stage beam search module is used to perform a two-stage beam search on the candidate decision set using the improved Beam-ACO algorithm, expand and prune partial solutions, and retain the set of partial solutions with the best scores according to the preset beam width; The monitoring-triggered update module is used to perform hole sequence selection and segmented trajectory selection in a partial solution set, collect online monitoring data and perform threshold determination, and output hole sequence planning results and segmented trajectory planning results.

[0014] The trajectory output module is used to generate an executable scanning trajectory and platform motion path based on the hole sequence planning results, segmented trajectory planning results, and input parameter set.

[0015] The beneficial effects of this invention are: This invention unifies the modeling of the task parameters, kinematic constraint parameters, and process constraint parameters, constructs a trajectory template set, a candidate set of process parameters, and a candidate decision set, and introduces an improved Beam-ACO for optimization, enabling hole sequence planning and depth segment trajectory planning to be completed collaboratively within the same framework. This avoids the insufficient executability problem caused by the separate planning of hole sequence and single hole trajectory in the prior art, and improves the overall feasibility and stability of the trajectory planning results for deep hole laser drilling in glass.

[0016] This invention establishes a comprehensive cost evaluation system, which not only focuses on processing efficiency during the optimization process, but also constrains and suppresses risks. This ensures that the obtained trajectory planning results can effectively improve the quality fluctuation problem in deep hole processing while guaranteeing the processing cycle time, thereby improving hole shape consistency and processing reliability.

[0017] This invention employs an improved Beam-ACO algorithm with bandwidth control during the optimization process. By expanding, scoring, and pruning partial solutions to retain the optimal solution set, the search space size is significantly reduced while ensuring search quality. This improves the planning efficiency for laser drilling tasks involving multi-hole and large-scale glass deep holes, and avoids the problems of slow convergence speed and high computational cost that traditional ant colony algorithms tend to encounter in complex decision-making scenarios.

[0018] This invention introduces an online monitoring data acquisition channel and trigger threshold. During the processing, the comprehensive cost evaluation system can be dynamically updated based on the monitoring results, and the current partial solution set can be re-evaluated and re-optimized, thereby achieving adaptive adjustment of trajectory planning. This enables the system to respond to changes in processing status in a timely manner, further improving the stability, robustness and overall production efficiency of deep hole laser drilling in glass. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 The flowchart shows the glass deep hole laser drilling trajectory planning method and system based on ant colony algorithm proposed in this invention. Figure 2 This is a schematic diagram of the module of the glass deep hole laser drilling trajectory planning system based on ant colony algorithm proposed in this invention; Figure 3This is a schematic diagram of the core algorithm structure of the glass deep hole laser drilling trajectory planning method based on ant colony algorithm proposed in this invention. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0021] refer to Figure 1-3 A method and system for planning laser drilling trajectory for deep holes in glass based on ant colony algorithm, comprising the following steps: Obtain the task parameters, kinematic constraint parameters, and process constraint parameters of the glass workpiece as the input parameter set; Each hole to be processed is segmented into depths according to the parameters of the task to be processed, and a set of trajectory templates, a set of candidate process parameters, an online monitoring data acquisition channel and a trigger threshold are set. A candidate decision set is generated based on the parameters of the task to be processed, depth segmentation, trajectory template set, and candidate set of process parameters. Based on the candidate decision set, initialize the pheromone distribution and heuristic information, and establish a comprehensive cost evaluation system; An improved Beam-ACO algorithm is used on the candidate decision set to perform a two-level beam search of aperture order beam search and depth segmented trajectory beam search. At each decision level, the partial solutions formed by aperture order selection are expanded and pruned, and the set of partial solutions with the best scores is retained according to the preset beam width. In a partial solution set, hole sequence selection and segmented trajectory selection are performed. Online monitoring data is collected and threshold judgment is performed. When the judgment is valid, the comprehensive cost evaluation system is updated to obtain hole sequence planning results and segmented trajectory planning results. Based on the hole sequence planning results, segmented trajectory planning results, and input parameter set, an executable scanning trajectory and platform motion path are generated.

[0022] In this embodiment, obtaining the input parameter set specifically includes: The system receives the processing task parameters of the glass workpiece to be processed and writes them into the task data structure. The task data structure includes the number of holes, hole position coordinates, target hole diameter, minimum spacing between holes, and target hole depth. The number of holes and the number of hole position coordinate entries are checked to be consistent. The target hole diameter and target hole depth are checked one by one and are positive. The task data that passes the check is locked and saved. Receive the kinematic constraint parameters of the glass workpiece to be processed, including the maximum scanning speed of the galvanometer, the maximum scanning acceleration of the galvanometer, the maximum moving speed of the platform, the maximum moving acceleration of the platform, and the upper and lower limits of the Z-axis focusing stroke. Establish an association index between the kinematic constraint parameters and the task data structure and save it. The system receives process constraint parameters for the glass workpiece to be processed, including minimum dwell time, average laser power range, and pulse repetition frequency range. It then merges these process constraint parameters with the task data structure and kinematic constraint parameters to generate an input parameter set.

[0023] In this embodiment, the establishment of the trajectory template set, the process parameter candidate set, the online monitoring data acquisition channel, and the setting of the trigger threshold specifically include: Based on the parameters of the task to be processed, the target hole depth is read one by one, the segmented depth interval is set, and boundary points are generated from zero depth according to the segmented depth interval. Adjacent boundary points are paired to generate depth segments. The depth segments are numbered according to the generation order and the start and end depths of each segment are recorded. The depth segmentation results are output. Based on the depth segmentation results, a trajectory template is generated sequentially for each depth segment. The trajectory templates are circular scanning trajectory templates, spiral scanning trajectory templates, and intermittent slag removal scanning trajectory templates. Template numbers and adjustable parameter bits are written into each template. The trajectory templates are bound to the corresponding depth segments and stored. A set of trajectory templates is output. Based on the depth segmentation results, candidate parameter values ​​are generated sequentially for each depth segment and combined to form a candidate parameter group. For each candidate parameter group, the input parameter set is called to perform verification and remove candidate parameter groups that fail the verification. An online monitoring data acquisition channel is constructed and a trigger threshold is configured. The trigger threshold is associated with and stored with the corresponding depth segment. The online monitoring data acquisition channel includes a coaxial image acquisition channel, a plasma radiation intensity acquisition channel, and an acoustic emission acquisition channel. The trigger threshold includes an occlusion rate threshold, a fluctuation coefficient threshold, and an event energy threshold.

[0024] In this embodiment, the generation of the candidate decision set specifically includes: Assign a hole number to each hole to be processed, traverse any two holes and calculate the distance between them, compare the distance between holes with the minimum spacing between holes, and write the hole pairs that meet the spacing requirements into the hole sequence candidate table. The hole sequence candidate table records the starting hole number and the target hole number. Read the depth segmentation results hole by hole and segment by segment, read the trajectory template set and process parameter candidate set bound to each depth segment, pair each trajectory template and each candidate parameter group to generate depth segment candidate pairs, and write them into the depth segment candidate table. The depth segment candidate table records the hole number, segment number, trajectory template identifier and candidate parameter group identifier. Traverse each hole sequence transfer candidate in the hole sequence candidate table, read the starting hole number and the target hole number, traverse the depth segment candidate table entries corresponding to the target hole and read the segment number, trajectory template identifier and candidate parameter group identifier, and combine the hole sequence transfer candidate with the segment number, trajectory template identifier and candidate parameter group identifier to obtain the candidate decision set.

[0025] In this embodiment, obtaining the comprehensive cost evaluation system specifically includes: Read the candidate decision set table and iterate through each one, assign a candidate decision number to each candidate decision, write it into the candidate index table, and output the candidate index table. The pheromone distribution is initialized based on the candidate index table, the pore order pheromone table and the segmented pheromone table are established, and a unified initial pheromone value is written for each pore order transition candidate and candidate decision. The pore order pheromone table and the segmented pheromone table are then output. For each candidate decision, the processing time cost and kinematic constraint penalty are calculated and written sequentially. When calculating the processing time cost, the hole position coordinates are read and the straight-line distance between holes is calculated. The maximum moving speed of the platform is read and the transfer time is calculated accordingly. The trajectory template and candidate parameter group are read and the scanning path length and scanning speed are obtained and the scanning time is calculated accordingly. The minimum dwell time is read and written and then summed to obtain the processing time cost. When calculating the kinematic constraint penalty, the speed requirement and acceleration requirement corresponding to the candidate parameter group are compared with the kinematic constraint penalty respectively, and the excess part is cumulatively penalized according to the preset penalty coefficient. After completing the writing of each cost component, the cost components are weighted and summed according to the preset weight to obtain the comprehensive cost value. The comprehensive cost value is then processed by taking the reciprocal and superimposing a preset positive number to obtain the heuristic information value and written into the heuristic information table. The comprehensive cost evaluation system is then output.

[0026] This invention establishes a dual pheromone distribution of hole sequence pheromone and segmented pheromone, and generates heuristic information with a comprehensive cost consisting of processing time cost and kinematic constraint penalty to guide the selection and iterative optimization of candidate decisions. This makes the planning process more likely to converge to an executable solution under constraints such as speed and acceleration, reduces invalid searches and manual corrections, and improves the computational efficiency, trajectory feasibility and processing cycle stability of laser drilling trajectory planning for deep holes in glass.

[0027] In this embodiment, obtaining the set of partial solutions with the optimal score specifically includes: An improved Beam-ACO algorithm is used on the candidate decision set. The candidate decision set, pheromone distribution and heuristic information are read, a partial solution data structure is established and empty hole sequence and empty segmented trajectory are written. The current decision layer is set as the starting layer. The partial solution containing only empty hole sequence and empty segmented trajectory is written into the partial solution set and output as the initial partial solution set. The initial partial solution set is copied into an ant construction queue corresponding to the number of ants and the current partial solution set of each ant is initialized. The ant is an independent solution construction instance used to construct candidate hole sequence and segmented trajectory. Each ant corresponds to a hole sequence selection sequence and a set of segmented trajectory selection sequences. The hole order search is performed using a two-level beam search. The initial partial solution set of the current decision layer is traversed one by one in an ant-constructed queue. For each partial solution of each ant, the selected hole set and the current last hole number are read. Hole order candidates with the starting hole number equal to the current last hole number and the target hole number not appearing in the selected hole set are selected from the candidate decision set. For each hole order candidate, the corresponding pheromone value and the corresponding heuristic information value are read. The selection weight of the hole order candidate is calculated according to the pheromone value and the heuristic information value, and all selection weights are normalized to obtain the selection probability. Sampling is performed according to the selection probability or the target hole number is determined according to the maximum probability rule. The target hole number is appended to the partial solution to form the hole order extended partial solution. When forming the hole order extended partial solution, the transfer time is included in the processing time cost. Then the hole order extended partial solution is written into the extended set. A two-stage beam search for depth-segmented trajectory is performed, iterating through the extended set line by line. For each hole sequence extension, the newly added target hole number is read and the depth segmentation result of the target hole is read. Segmented trajectory candidates matching the target hole number and depth segment number are sequentially selected from the candidate decision set according to the depth segment number. Feasibility checks are performed on each segmented trajectory candidate to eliminate candidates that violate kinematic and process constraints. For candidates that fail the check but are repairable, repair operations are performed, including inserting a dwell cooling, reducing the scanning speed, switching to an intermittent slag-removing scanning trajectory template, and correcting... After verification or repair, the incremental cost of each segmented trajectory candidate is calculated and added to the determined cost of the partial solution to form a segmented expansion score. The incremental cost includes the scanning time and dwell cooling time of the current depth segment and is included in the processing time cost, and includes the kinematic constraint penalty increment obtained by accumulating the speed and acceleration requirements corresponding to the candidate parameter group that exceed the galvanometer or platform constraints and is included in the kinematic constraint penalty. Segmented trajectory candidates for each depth segment are selected from best to worst according to the segmented expansion score and combined to form different segmented expansion partial solutions. The segmented expansion partial solutions are written into the score expansion set. The score extension set is sorted from best to worst and pruned. The score is obtained by weighting the processing time cost and kinematic constraint penalty. The pruning operation is to retain only the partial solutions with the best score, which does not exceed the preset bundle width, as the partial solution set of the next decision layer. The pheromone distribution is updated. The aperture order pheromone and segmented pheromone are proportionally attenuated according to the pheromone evaporation parameter. The aperture order selection and segmented trajectory selection that appear in the partial solutions with the best score are written with pheromone increment parameters. The partial solution set with the best score is output.

[0028] This invention addresses the problems of strong coupling between multi-hole sequence and single-hole segmented trajectory in laser drilling of deep glass holes. These issues, coupled with a huge candidate space and strict constraints imposed by kinematics and process limitations, lead to the tendency of traditional ant colony algorithms to blindly expand, converge slowly, and generate unexecutable solutions. An improved Beam-ACO approach is employed to explicitly decompose solution construction into a two-stage beam search: a hole sequence beam search and a segmented trajectory beam search. Beam width pruning is used to retain only the optimal partial solution set at each decision level, compressing the search tree width and improving convergence efficiency. Furthermore, pheromones and heuristic information are used to guide the hole sequence transition probability. The rate is increased and the transfer time is included in the cost to take into account the cycle time. Further, the feasibility of candidates is checked at the segmented trajectory layer, and repairable candidates are repaired by means of resident cooling insertion, scanning speed reduction or switching of intermittent slag removal template to improve the density of feasible solutions and reduce rework. Then, the incremental cost is accumulated to form a comparable segmented expansion score and connect it with the hole sequence expansion. Finally, the stable exploration and rapid utilization are achieved by combining pheromone volatilization and incremental reinforcement of the optimal partial solution. Thus, under complex constraints, executable hole sequence and segmented trajectory planning results are obtained faster and the planning quality and stability are improved.

[0029] In this embodiment, obtaining the hole sequence planning result and the segmented trajectory planning result specifically includes: Read a partial solution set and use it as the current candidate set. Perform hole order selection and segmented trajectory selection on the current candidate set. The hole order selection calculates the selection weight based on the hole order pheromone and hole order heuristic information and determines the next target hole number, which is written into the hole order field. The segmented trajectory selection filters feasible candidates for each depth segment of the target hole and forms a target vector. At the same time, non-dominated sorting is performed to extract non-dominated solutions. When the limit is exceeded, the solution is filtered by the crowding distance and written into the segmented trajectory field. Output the hole order candidate set. The candidate set of hole order is determined with the corresponding trigger threshold. If any trigger threshold is exceeded, a trigger mark is generated. The comprehensive cost evaluation system is updated for the piecewise extended partial solution with the trigger mark, and the processing time cost and kinematic constraint penalty of the piecewise extended partial solution are updated simultaneously. The sorting and pruning are re-executed and the hole order field and piecewise trajectory field are read to obtain the hole order planning result and the piecewise trajectory planning result. In this embodiment, the generation of the executable scanning trajectory and platform motion path specifically includes: Read the hole sequence planning results, segmented trajectory planning results and input parameter set, write them together into the generated execution plan data and archive it. The plan data includes the hole sequence, hole position coordinates, template identifiers and parameter group identifiers for each hole and segment. Read the hole sequence and hole position coordinates from the execution plan data, set the initial position of the platform as the path start point and write it into the path point sequence, write the hole position coordinates into the path point sequence as the platform target point, calculate the transfer distance between adjacent points in the path point sequence segment by segment, write the start point, end point, transfer distance and transfer time of each transfer segment into the platform motion segment record in sequence, and output the platform path and scheduling table. Read and verify the process constraint parameters and kinematic constraint parameters one hole at a time and segment at a time, generate the scan trajectory point sequence and encapsulation sequence according to the template, and export the scan trajectory file and platform path file.

[0030] A laser drilling trajectory planning system for deep holes in glass based on ant colony algorithm includes: The parameter acquisition module is used to acquire the processing parameters, kinematic constraint parameters, and process constraint parameters of the glass workpiece. The segmented database module is used to establish a set of trajectory templates and a set of candidate process parameters, as well as to establish an online monitoring data acquisition channel and set trigger thresholds. The candidate generation module is used to generate a candidate decision set based on the parameters of the task to be processed, depth segmentation, trajectory template set, and process parameter candidate set. The evaluation initialization module is used to initialize the pheromone distribution and heuristic information based on the candidate decision set, and to establish a comprehensive cost evaluation system. The two-stage beam search module is used to perform a two-stage beam search on the candidate decision set using the improved Beam-ACO algorithm, expand and prune partial solutions, and retain the set of partial solutions with the best scores according to the preset beam width; The monitoring-triggered update module is used to perform hole sequence selection and segmented trajectory selection in a partial solution set, collect online monitoring data and perform threshold determination, and output hole sequence planning results and segmented trajectory planning results.

[0031] The trajectory output module is used to generate an executable scanning trajectory and platform motion path based on the hole sequence planning results, segmented trajectory planning results, and input parameter set.

[0032] Example 1: To verify the feasibility of this invention in practice, it is applied to a typical glass deep-hole laser drilling scenario. This scenario addresses the need for machining deep-hole arrays on hard and brittle glass workpieces, characterized by numerous holes, small diameters, and large depths. The large number and dense distribution of holes place high demands on the platform's motion path, galvanometer scanning trajectory, machining cycle time, and stability. In this type of machining task, if the hole sequence planning is unreasonable or the scanning trajectory and parameter settings for individual holes vary with depth, problems such as excessively long platform idle travel time, unexecutable scanning trajectories under speed and acceleration constraints, poor slag removal leading to occlusion during deep-hole machining, and frequent manual intervention to adjust parameters can easily occur, severely impacting overall machining efficiency and stability.

[0033] In this scenario, the hole distribution information, target hole diameter, and target hole depth of the glass workpiece are first acquired, along with the kinematic and process constraint parameters of the laser processing equipment, serving as the input parameter set for this invention. For each hole to be processed, the deep hole processing process is divided into multiple depth segments based on the target hole depth. Adaptive scanning trajectory templates and candidate sets of process parameters are established for each depth segment, enabling different scanning methods and parameter combinations to be used for different depth segments to meet the processing characteristics that vary with depth during deep hole processing. Simultaneously, coaxial vision and acoustic emission online monitoring data acquisition channels are configured in the processing system, and trigger thresholds are set for the monitoring indicators to reflect any abnormal states that may occur during processing.

[0034] In the trajectory planning phase, based on the aforementioned depth segmentation results, trajectory template set, and candidate set of process parameters, a candidate decision set is constructed, encompassing hole sequence transfer decisions and segmented trajectory decisions. Furthermore, a comprehensive cost evaluation system is established. This system focuses on processing time cost and kinematic constraint penalties, ensuring that the optimization process not only considers whether the hole sequence and trajectory are "shortest," but also constrains parameters such as scanning speed, acceleration, and dwell cooling to remain within the allowable ranges of the equipment and process. Building upon this, an improved Beam-ACO algorithm is used to search the candidate decision set. Through joint optimization of hole sequence bundle search and depth segmented trajectory bundle search, only the optimal solution set is retained for subsequent expansion, effectively controlling the search scale.

[0035] In practical applications, compared to the traditional method of setting hole sequence and fixed scanning trajectory based on experience, this invention can automatically generate a combination of hole sequence and segmented trajectory that meets equipment constraints and has a shorter overall processing time. For the same glass workpiece processing task, the traditional method usually requires process engineers to perform multiple trial cuts and manually adjust the hole sequence and parameters, which is time-consuming in the planning stage, and there may still be situations where local speed exceeds the limit or temporary cooling is required during execution. This invention can take these constraints into account in the planning stage and repair or eliminate infeasible candidates during the search process, so that the final generated trajectory can be executed directly.

[0036] To verify the beneficial effects, the performance of the method of this invention and the traditional experience-based planning method were compared under the same processing task. The traditional experience-based planning method involves process engineers manually determining the hole processing sequence based on past processing experience, and generating platform paths and scanning trajectories for each hole using preset fixed scanning trajectories and fixed or slightly adjusted process parameters. During execution, manual corrections are made based on processing phenomena. Specific comparison data are shown in Table 1: Table 1. Comparison of the Invention Method and Traditional Methods

[0037] As shown in Table 1, compared with the traditional solution, the present invention achieves significant improvements in both planning and execution: trajectory planning time is reduced from 180 s to 65 s, the total platform movement distance is reduced from 12.4 m to 9.1 m, and the overall processing time is reduced from 42 min to 31 min; at the same time, executability is significantly improved, the number of unexecutable trajectories is reduced from 5 times / task to 0 times / task, the number of manual intervention adjustments is reduced from 4 times / task to 1 time / task, and the number of additional dwell cooling insertions is reduced from 9 times / task to 2 times / task; in terms of processing quality, the first-pass yield rate is increased from 83.3% to 98.9%, the hole edge chipping rate is reduced from 3.8% to 1.2%, and the hole wall roughness Ra is reduced from 1.35 µm to 0.92 µm, indicating that the present invention improves efficiency and trajectory feasibility while further enhancing the stability and hole shape consistency of deep hole processing.

[0038] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for planning the trajectory of laser drilling for deep holes in glass based on ant colony algorithm, characterized in that, Includes the following steps: Obtain the task parameters, kinematic constraint parameters, and process constraint parameters of the glass workpiece as the input parameter set; Each hole to be processed is segmented into depths according to the parameters of the task to be processed, and a set of trajectory templates, a set of candidate process parameters, an online monitoring data acquisition channel and a trigger threshold are set. A candidate decision set is generated based on the parameters of the task to be processed, depth segmentation, trajectory template set, and candidate set of process parameters. Based on the candidate decision set, initialize the pheromone distribution and heuristic information, and establish a comprehensive cost evaluation system; An improved Beam-ACO algorithm is used on the candidate decision set to perform a two-level beam search of aperture order beam search and depth segmented trajectory beam search. At each decision level, the partial solutions formed by aperture order selection are expanded and pruned, and the set of partial solutions with the best scores is retained according to the preset beam width. In a partial solution set, hole sequence selection and segmented trajectory selection are performed. Online monitoring data is collected and threshold judgment is performed. When the judgment is valid, the comprehensive cost evaluation system is updated to obtain hole sequence planning results and segmented trajectory planning results. Based on the hole sequence planning results, segmented trajectory planning results, and input parameter set, an executable scanning trajectory and platform motion path are generated.

2. The method for planning laser drilling trajectory for deep holes in glass based on ant colony algorithm according to claim 1, characterized in that, The specific steps to obtain the input parameter set include: The system receives the processing task parameters of the glass workpiece to be processed and writes them into the task data structure. The task data structure includes the number of holes, hole position coordinates, target hole diameter, minimum spacing between holes, and target hole depth. The number of holes and the number of hole position coordinate entries are checked to be consistent. The target hole diameter and target hole depth are checked one by one and are positive. The task data that passes the check is locked and saved. Receive the kinematic constraint parameters of the glass workpiece to be processed, establish an association index between the kinematic constraint parameters and the task data structure, and save it; Receive the process constraint parameters of the glass workpiece to be processed, and merge the process constraint parameters with the task data structure and kinematic constraint parameters to generate an input parameter set.

3. The method for planning laser drilling trajectory for deep holes in glass based on ant colony algorithm according to claim 1, characterized in that, The establishment of the trajectory template set, the candidate set of process parameters, the online monitoring data acquisition channel, and the setting of the trigger threshold specifically include: Based on the parameters of the task to be processed, the target hole depth is read one by one, the segmented depth interval is set, and boundary points are generated from zero depth according to the segmented depth interval. Adjacent boundary points are paired to generate depth segments. The depth segments are numbered according to the generation order and the start and end depths of each segment are recorded. The depth segmentation results are output. Based on the depth segmentation results, a trajectory template is generated sequentially for each depth segment. The template number and adjustable parameter bits are written into each template. The trajectory template is then bound to the corresponding depth segment and stored. A set of trajectory templates is output. Based on the depth segmentation results, candidate parameter values ​​are generated sequentially for each depth segment and combined to form a candidate parameter group. For each candidate parameter group, the input parameter set is called to perform verification and remove candidate parameter groups that fail the verification. An online monitoring data acquisition channel is constructed and a trigger threshold is configured. The trigger threshold is associated with and stored with the corresponding depth segment.

4. The method for planning laser drilling trajectory for deep holes in glass based on ant colony algorithm according to claim 1, characterized in that, The generation of the candidate decision set specifically includes: Assign a hole number to each hole to be processed, traverse any two holes and calculate the distance between them, compare the distance between holes with the minimum spacing between holes, and write the hole pairs that meet the spacing requirements into the hole sequence candidate table. Read the depth segmentation results hole by hole and segment by segment, read the trajectory template set and process parameter candidate set bound to each depth segment, pair each trajectory template and each candidate parameter group to generate depth segment candidate pairs, and write them into the depth segment candidate table. Traverse each hole sequence transfer candidate in the hole sequence candidate table, read the starting hole number and the target hole number, traverse the depth segment candidate table entries corresponding to the target hole and read the segment number, trajectory template identifier and candidate parameter group identifier, and combine the hole sequence transfer candidate with the segment number, trajectory template identifier and candidate parameter group identifier to obtain the candidate decision set.

5. The method for planning laser drilling trajectory for deep holes in glass based on ant colony algorithm according to claim 1, characterized in that, The comprehensive cost evaluation system is obtained by specifically including: Read the candidate decision set table and iterate through each one, assign a candidate decision number to each candidate decision, write it into the candidate index table, and output the candidate index table. The pheromone distribution is initialized based on the candidate index table, the pore order pheromone table and the segmented pheromone table are established, and a unified initial pheromone value is written for each pore order transition candidate and candidate decision. The pore order pheromone table and the segmented pheromone table are then output. For each candidate decision, the processing time cost and kinematic constraint penalty are calculated and written in sequence. After writing each cost component, the cost components are weighted and summed according to the preset weights to obtain the comprehensive cost value. The comprehensive cost value is then processed by taking the reciprocal and adding a preset positive number to obtain the heuristic information value, which is written into the heuristic information table. The comprehensive cost evaluation system is then output.

6. The method for planning laser drilling trajectory for deep holes in glass based on ant colony algorithm according to claim 1, characterized in that, The specific steps involved in obtaining the set of partial solutions with the optimal score are as follows: An improved Beam-ACO algorithm is used on the candidate decision set. The candidate decision set, pheromone distribution and heuristic information are read, a partial solution data structure is established and empty hole sequence and empty segment trajectory are written. The current decision layer is set as the starting layer. The partial solution containing only empty hole sequence and empty segment trajectory is written into the partial solution set and output as the initial partial solution set. The initial partial solution set is copied to the ant construction queue corresponding to the number of ants and the current partial solution set of each ant is initialized. The hole order search is performed using a two-level beam search. The initial partial solution set of the current decision layer is traversed one by one in an ant-constructed queue. For each partial solution of each ant, the selected hole set and the current last hole number are read. Hole order candidates with the starting hole number equal to the current last hole number and the target hole number not appearing in the selected hole set are selected from the candidate decision set. For each hole order candidate, the corresponding pheromone value and the corresponding heuristic information value are read. The selection weight of the hole order candidate is calculated according to the pheromone value and the heuristic information value, and all selection weights are normalized to obtain the selection probability. Sampling is performed according to the selection probability or the target hole number is determined according to the maximum probability rule. The target hole number is appended to the partial solution to form the hole order extended partial solution. When forming the hole order extended partial solution, the transfer time is included in the processing time cost. Then the hole order extended partial solution is written into the extended set. The depth segmented trajectory bundle search is performed using a two-stage bundle search. The extended set is traversed one by one. For each hole sequence extension partial solution, the newly added target hole number and the depth segmentation result of the target hole are read. Segmented trajectory candidates that match the target hole number and depth segment number are selected from the candidate decision set according to the depth segment number. Feasibility verification is performed on each segmented trajectory candidate to eliminate candidates that violate kinematic constraint parameters and process constraint parameters. For candidates that fail the verification but can be repaired, repair operations are performed. For segmented trajectory candidates that pass the verification or are repaired, the incremental cost is calculated one by one and added to the determined cost of the partial solution to form a segmented extension score. Segmented trajectory candidates for each depth segment are selected from best to worst according to the segmented extension score and combined to form different segmented extension partial solutions. The segmented extension partial solutions are written into the score extension set. The score extension set is sorted from best to worst and pruned. The score is obtained by weighting the processing time cost and kinematic constraint penalty. An update operation is performed on the pheromone distribution. The pore order pheromone and segmented pheromone are proportionally attenuated according to the pheromone evaporation parameter. The pore order selection and segmented trajectory selection that appear in the partial solution with the best score are written with pheromone increment parameter. The set of partial solutions with the best score is output.

7. The method for planning laser drilling trajectory for deep holes in glass based on ant colony algorithm according to claim 1, characterized in that, The results of the hole sequence planning and the piecewise trajectory planning are obtained specifically in the following ways: Read a partial solution set and use it as the current candidate set. Perform hole order selection and segmented trajectory selection on the current candidate set. At the same time, perform non-dominated sorting to extract non-dominated solutions. If the limit is exceeded, filter by crowding distance and write it into the segmented trajectory field. Output the hole order candidate set. The candidate set of hole order is determined by the corresponding trigger threshold. If any trigger threshold is exceeded, a trigger mark is generated. The comprehensive cost evaluation system is updated for the piecewise extended partial solution with the trigger mark, and the processing time cost and kinematic constraint penalty of the piecewise extended partial solution are updated simultaneously. The sorting and pruning are re-executed, and the hole order field and piecewise trajectory field are read to obtain the hole order planning result and the piecewise trajectory planning result.

8. The method for planning laser drilling trajectory for deep holes in glass based on ant colony algorithm according to claim 1, characterized in that, The generation of the executable scanning trajectory and platform motion path specifically includes: Read the hole sequence planning results, segmented trajectory planning results and input parameter set, write them together into the generated execution plan data and archive them; Read the hole sequence and hole position coordinates from the execution plan data, set the initial position of the platform as the path start point and write it into the path point sequence, write the hole position coordinates into the path point sequence as the platform target point, calculate the transfer distance between adjacent points in the path point sequence segment by segment, write the start point, end point, transfer distance and transfer time of each transfer segment into the platform motion segment record in sequence, and output the platform path and scheduling table. Read and verify the process constraint parameters and kinematic constraint parameters one hole at a time and segment at a time, generate the scan trajectory point sequence and encapsulation sequence according to the template, and export the scan trajectory file and platform path file.

9. A glass deep hole laser drilling trajectory planning system based on ant colony algorithm, comprising executing the glass deep hole laser drilling trajectory planning method based on ant colony algorithm as described in any one of claims 1 to 8, characterized in that, include: The parameter acquisition module is used to acquire the processing parameters, kinematic constraint parameters, and process constraint parameters of the glass workpiece. The segmented database module is used to establish a set of trajectory templates and a set of candidate process parameters, as well as to establish an online monitoring data acquisition channel and set trigger thresholds. The candidate generation module is used to generate a candidate decision set based on the parameters of the task to be processed, depth segmentation, trajectory template set, and process parameter candidate set. The evaluation initialization module is used to initialize the pheromone distribution and heuristic information based on the candidate decision set, and to establish a comprehensive cost evaluation system. The two-stage beam search module is used to perform a two-stage beam search on the candidate decision set using the improved Beam-ACO algorithm, expand and prune partial solutions, and retain the set of partial solutions with the best scores according to the preset beam width; The monitoring-triggered update module is used to perform hole sequence selection and segmented trajectory selection in a partial solution set, collect online monitoring data and perform threshold determination, and output hole sequence planning results and segmented trajectory planning results. The trajectory output module is used to generate an executable scanning trajectory and platform motion path based on the hole sequence planning results, segmented trajectory planning results, and input parameter set.

Images