Machine learning based intelligent dispatch management system for PCB drilling machine

Abstract

This invention discloses a machine learning-based intelligent dispatch management system for PCB drilling machines, comprising the following modules: a process deconstruction module for generating a PCB drilling task process set; a capability building module for constructing a real-time capability set for the drilling machine; a segment division module for forming a task tool execution segment table; an adaptation calculation module for processing using an improved Resonance model to obtain a task-equipment adaptation list; a sequence arrangement module for generating a drilling machine sequential dispatch table; a position locking module for generating a machine position dispatch instruction set; a disturbance optimization module for updating operating status data and generating a disturbance-corrected dispatch sequence based on a penetration-driven Jaya scheduling algorithm; and a handover generation module for generating a continuous processing handover plan. This invention combines an improved Resonance model and a penetration-driven Jaya scheduling algorithm to achieve intelligent and efficient dispatching of PCB drilling machines.

Description

Technical Field

[0001] This invention relates to the field of intelligent scheduling technology for drilling machines, and more particularly to an intelligent dispatch management system for PCB drilling machines based on machine learning. Background Technology

[0002] As the electronics manufacturing industry continues to develop towards higher density, higher precision, and more diverse, smaller-batch production, the demands for intelligent and refined production scheduling in PCB drilling are increasing. Efficient matching and dynamic dispatching management technologies between various drilling tasks and multiple drilling machines have attracted widespread attention. Existing PCB drilling machine dispatching systems mainly rely on rule-based static allocation methods or simple priority ranking strategies for task assignment. However, these systems commonly suffer from the following problems in practical applications: Existing task assignment methods typically rely solely on order priority or aperture type for coarse-grained task division, lacking comprehensive consideration of process characteristics. This leads to unreasonable task allocation and discontinuous processing paths, increasing idle travel distance and tool change frequency, and reducing overall processing efficiency. Regarding equipment capacity matching, existing methods often use static equipment parameters or single state indicators for judgment, failing to accurately reflect dynamic operating characteristics such as spindle vibration, temperature changes, and machine occupancy status. This results in high-load equipment still being assigned high-intensity tasks, exacerbating equipment vibration, accelerating tool wear, and reducing processing accuracy. In the task-equipment adaptation calculation process, traditional methods rely heavily on simple feature similarity or empirical rules, lacking the ability to model the coupling relationship between multi-dimensional process characteristics and multi-dimensional equipment responses, making accurate matching under complex working conditions difficult. Furthermore, in terms of task assignment result optimization, existing scheduling strategies typically lack dynamic disturbance correction, failing to adjust the task order in real time based on equipment status changes during execution. This easily leads to local optima or even scheduling failures, resulting in poor production continuity, frequent machine changes, and unstable overall production efficiency.

[0003] Therefore, how to provide a machine learning-based intelligent dispatch management system for PCB drilling machines is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] One objective of this invention is to propose an intelligent dispatch management system for PCB drilling machines based on machine learning. This invention combines an improved Resonance model with a penetration-driven Jaya scheduling algorithm. By constructing a PCB drilling task process set and a drilling machine real-time capability set, it uses a process spectrum generation unit and a device response spectrum generation unit to form task process spectrum sequences and device response spectrum sequences. Furthermore, a temporal harmonic window penetration mechanism is introduced in the harmonic window penetration unit to perform temporal coupling and nonlinear transformation processing on the non-negative response matrix, obtaining the harmonic window penetration sequence and normalized contribution value, thus achieving accurate generation of the task-equipment adaptation list. Based on this, by combining machine station occupancy locking operation and the penetration-driven Jaya scheduling algorithm, local rearrangement and dynamic optimization are performed on the adaptation execution segment to generate a disturbance-corrected dispatch sequence and form a continuous processing handover plan, thereby improving the matching accuracy, processing continuity, and overall production efficiency of PCB drilling dispatch.

[0005] The intelligent dispatch management system for PCB drilling machines based on machine learning according to an embodiment of the present invention includes the following modules: The process deconstruction module is used to receive PCB orders to be produced and corresponding drilling process files, deconstruct the process of each PCB drilling task, and generate a PCB drilling task process set. The capability building module is used to collect the operating status data of each PCB drilling machine and build a real-time capability set for the drilling machine. The segment division module is used to divide the tool execution segments of each PCB drilling task on each PCB drilling machine based on the PCB drilling task process set and the drilling machine real-time capability set, forming a task tool execution segment table. The adaptation calculation module is used to input the task tool execution segment table and the drilling machine real-time capability set into the improved Resonance model for processing to obtain the task-equipment adaptation list. The improved Resonance model includes a process spectrum generation unit, an equipment response spectrum generation unit, a harmonic window penetration unit, and a dispatch adaptation output unit. The harmonic window penetration unit has a built-in time-series harmonic window penetration mechanism. The sequence arrangement module is used to arrange the PCB drilling tasks in sequence based on the task-equipment adaptation list and generate a drilling machine sequence dispatch table. The occupancy locking module is used to perform machine occupancy locking operations on each PCB drilling machine based on the drilling machine sequence dispatch table, and generate a set of machine occupancy dispatch instructions. The disturbance optimization module is used to update the operating status data of each PCB drilling machine during the execution of the machine occupancy dispatch instruction set, and perform local rearrangement based on the penetration-driven Jaya scheduling algorithm to generate the disturbance correction dispatch sequence for the corresponding PCB drilling machine. The handover generation module is used to take over the next stage of tasks for PCB drilling machines that have completed the current processing segment based on the disturbance correction dispatch sequence, and generate a continuous processing handover plan.

[0006] Optionally, the process deconstruction module specifically comprises: Receive PCB orders to be produced and corresponding drilling process documents, parse the drilling process documents, and extract hole position coordinate data, hole group distribution area data, board thickness segment data, hole diameter data, and hole sequence path data; Based on the aperture size data, aperture levels are divided to obtain aperture level data; The drilling path length is calculated based on the hole position coordinate data and hole sequence path data. The drilling path length is the cumulative distance between adjacent hole position coordinates. The number of path switching times is calculated based on the hole sequence path data, where the number of path switching times is the number of times adjacent processing directions change in the hole sequence path; The hole diameter level data, hole group distribution area data, board thickness section data, drilling path length, and path switching number are structured and combined to generate a PCB drilling task process set.

[0007] Optionally, the operating status data includes spindle speed data, spindle vibration amplitude data, spindle temperature data, current tool diameter data, and machine tool occupancy status data.

[0008] Optionally, the segmentation module specifically comprises: Based on the hole diameter level data, hole group distribution area data, board thickness segment data, drilling path length and path switching number in the PCB drilling task process set, each PCB drilling task is segmented according to the hole diameter level change position and hole sequence path order to obtain the initial processing segment set. Based on the current tool diameter data in the real-time capability set of the drilling machine, the hole diameter is matched for each processing segment in the initial processing segment set. The processing segment whose hole diameter level data is consistent with the current tool diameter data is mapped to the corresponding PCB drilling machine to obtain the first execution segment set. Based on the hole group distribution area data, each processing segment in the first execution segment set is grouped according to its hole group distribution area, and processing segments that are consecutive in the same hole group distribution area and in the hole sequence path are merged to obtain the second execution segment set. Based on the plate thickness segment data, each processing segment in the second execution segment set is divided according to the plate thickness segment, and processing segments with the same plate thickness segment are merged to obtain the third execution segment set; Based on the drilling path length and the number of path switching, the processing segments in the third execution segment set are rearranged according to the path order, and the processing segments with adjacent paths that do not increase the number of path switching after splicing are spliced ​​together to obtain the fourth execution segment set. Based on the spindle speed data, spindle vibration amplitude data, and spindle temperature data, the operating status of each processing segment in the fourth execution segment set is matched, and the processing segment corresponding to the operating status is marked as the matching segment to obtain the fifth execution segment set; Based on the machine occupancy status data, the occupancy status of each processing segment in the fifth execution segment set is determined, and the processing segments with idle machine occupancy status are retained to obtain the target execution segment set; The target execution segment set is combined according to the correspondence between PCB drilling tasks and PCB drilling machines to form a task tool execution segment table.

[0009] Optionally, the adaptation calculation module specifically comprises: The task tool execution segment table is input into the process spectrum generation unit. The target execution segment set corresponding to each PCB drilling task and PCB drilling machine in the task tool execution segment table is expanded. The hole diameter level data, hole group distribution area data, board thickness segment data, drilling path length and path switching number in each target execution segment are arranged in order to form a multi-dimensional process vector. The value of each dimension is divided by the maximum value of the corresponding dimension to generate the task process spectrum and construct the task process spectrum sequence. The PCB drilling machine correspondence in the task tool execution segment table is input into the device response spectrum generation unit. Based on the spindle speed data, spindle vibration amplitude data, spindle temperature data, current tool diameter data, and machine occupancy status data in the real-time capability set of the drilling machine, the operating status data of the PCB drilling machine corresponding to the target execution segment is extracted one by one. The operating status data are arranged into a multi-dimensional response vector in chronological order. The value of each dimension is divided by the maximum value of the corresponding dimension to generate the device response spectrum and construct the device response spectrum sequence. The task process spectrum sequence and the equipment response spectrum sequence are input into the harmonic window penetration unit. Each process spectrum in the task process spectrum sequence is subtracted from the corresponding PCB drilling machine's equipment response spectrum to generate a spectrum difference matrix. The values ​​less than zero in the spectrum difference matrix are set to zero to generate a non-negative response matrix. The timing harmonic window penetration mechanism is called on the non-negative response matrix for processing to generate a harmonic window penetration sequence. The harmonic window penetration sequence is input into the dispatch adapter output unit, and the harmonic window penetration sequence is subjected to forward accumulation operation according to the target execution segment order to generate the accumulated response value of the corresponding target execution segment. The cumulative response values ​​corresponding to all target execution segments are summed to obtain the total response. The cumulative response value corresponding to each target execution segment is divided by the total response to generate the normalized contribution value of the corresponding target execution segment. Select the target execution segment whose normalized contribution value is greater than or equal to the set contribution threshold as the adapted execution segment, and generate the task-device adaptation list.

[0010] Optionally, the timing harmonic window pass-through mechanism specifically includes: For each row of data in the non-negative response matrix, perform a dimension-wise subtraction operation with the adjacent previous row of data. When the current row is the first row, use the current row of data as the adjacent previous row of data for the operation to generate the forward original difference matrix. Set the values ​​less than zero in the forward original difference matrix to zero to generate the forward difference matrix. For each row of data in the non-negative response matrix, perform a dimension-wise subtraction operation with the adjacent next row of data. When the current row is the last row, treat the current row of data as the adjacent next row of data for the operation to generate the backward original difference matrix. Set the values ​​less than zero in the backward original difference matrix to zero to generate the backward difference matrix. The forward difference matrix and the backward difference matrix are multiplied to generate a bidirectional coupling matrix; Each element in the bidirectional coupling matrix is ​​incremented by 1 and subjected to a natural logarithm operation, wherein the natural logarithm operation is a logarithmic operation with the constant e as the base and the corresponding element value as the argument, thereby generating a logarithmic coupling matrix; Multiply each element in the logarithmic coupling matrix by the reciprocal of the corresponding target execution segment index value to generate a scaled logarithmic matrix; An exponential function operation is performed on each element in the scaled logarithmic matrix. The exponential function operation is a power operation with a constant e as the base and the corresponding element value as the exponent, to generate an exponential response matrix. Normalize the data in each row of the exponential response matrix by dividing each element in each row by the sum of all elements in the corresponding row to generate a normalized response matrix. Each row of data in the normalized response matrix is ​​multiplied dimension-by-dimensionally with the reciprocal of the corresponding target execution segment index value to generate a position-weighted response matrix; Each row of data in the position-weighted response matrix is ​​accumulated dimension by dimension to generate a penetration response value, which is then arranged according to the target execution segment order to generate a harmonic window penetration sequence.

[0011] Optionally, the sequential arrangement module specifically comprises: Based on the task-equipment adaptation list, the adaptation execution segments corresponding to each PCB drilling task are grouped according to the PCB drilling machine to generate a machine task set. In the set of machine tasks, the adaptation execution segments are grouped according to the aperture level data. Adaptive execution segments with the same aperture level data are grouped into the same group to generate an aperture group execution sequence. In the aperture grouping execution sequence, the adaptation execution segments are grouped according to the aperture distribution area data, and the adaptation execution segments with the same aperture distribution area data are divided into the same group to generate a region grouping execution sequence; In the aforementioned regional grouping execution sequence, the adaptation execution segments are grouped according to the plate thickness segment data. Adaptive execution segments with the same plate thickness segment data are grouped into the same group to generate a plate thickness grouping execution sequence. The adaptation execution segments within each group of the plate thickness grouping execution sequence are sorted from largest to smallest according to their corresponding normalized contribution values ​​to generate a sorted execution sequence within the group. The maximum normalized contribution value within each group is used as the inter-group ranking value, and the groups are sorted from largest to smallest according to the inter-group ranking value to generate an inter-group ranking sequence. Arrange the intra-group sorting execution sequences of each group according to the inter-group sorting sequence to form the task execution order of the corresponding PCB drilling machine. The task execution order of each PCB drilling machine is summarized to form a drilling machine sequence dispatch table.

[0012] Optionally, the placeholder locking module specifically comprises: Based on the drilling machine sequence dispatch table, the task execution order corresponding to each PCB drilling machine is extracted, and the corresponding appropriate execution segment is determined one by one according to the task execution order. For each PCB drilling machine, the adaptation execution segment is allocated a continuous time period according to the task execution order to generate a machine occupancy time sequence. Bind each time period in the machine occupancy time sequence to the corresponding adaptive execution segment to generate a machine occupancy mapping relationship; Based on the machine occupancy mapping relationship, a machine occupancy locking operation is performed on each PCB drilling machine. The PCB drilling machine corresponding to the time period within the machine occupancy time sequence is marked as locked; the PCB drilling machine corresponding to the time period not within the machine occupancy time sequence is marked as unlocked. The machine occupancy mapping relationship and machine occupancy lock status are combined to generate a machine occupancy dispatch instruction set.

[0013] Optionally, the disturbance optimization module specifically comprises: Real-time updates of the operating status data of each PCB drilling machine, generating and updating the real-time capability set of the drilling machine; Based on the machine position locking status in the machine position dispatch instruction set, the corresponding adaptation execution section is marked as a reserved section, and the corresponding adaptation execution section is marked as a rearrangement section. According to the task execution order in the drilling machine sequence dispatch table, extract the rearrangement segment corresponding to each PCB drilling machine and arrange them according to the task execution order to generate a local rearrangement segment sequence. For each PCB drilling machine, the penetration-driven Jaya scheduling algorithm is executed. Each rearranged segment in the local rearranged segment sequence is sequentially numbered according to the task execution order. The sequential number is used as the position number, and each rearranged segment is assigned a unique identifier. For each pair of adjacent rearranged segments in the local rearranged segment sequence, an adjacent exchange operation is executed once to generate multiple initial candidate dispatch sequences. The sum of contribution values ​​is obtained by summing the normalized contribution values ​​corresponding to each rearranged segment in the initial candidate dispatch sequence; the state penalty value is obtained by adding the normalized value of the spindle vibration amplitude data of the PCB drilling machine corresponding to the updated real-time capability set of the drilling machine to the normalized value of the spindle temperature data. Subtract the state penalty value from the sum of the contribution values ​​to obtain the sequence evaluation value. The initial candidate assignment sequence with the largest sequence evaluation value is determined as the optimal candidate assignment sequence, and the initial candidate assignment sequence with the smallest sequence evaluation value is determined as the worst candidate assignment sequence. Based on the unique identifier of each rearranged segment, the position number of the corresponding rearranged segment is found in the optimal candidate assignment sequence and the worst candidate assignment sequence, respectively. The difference between the current position number and the corresponding position number in the optimal candidate assignment sequence is calculated and the absolute value is taken to obtain the first position difference. The difference between the current position number and the corresponding position number in the worst candidate assignment sequence is calculated and the absolute value is taken to obtain the second position difference. The current position number is added to the first position difference and the second position difference is subtracted to obtain the updated position number. The rearranged segments are reordered according to the updated position number to generate an updated candidate assignment sequence, and the updated candidate assignment sequence is used as the initial candidate assignment sequence for the next iteration. Repeatedly perform sequence evaluation value calculation, optimal candidate assignment sequence determination, worst candidate assignment sequence determination, and update position number calculation until the maximum number of iterations or the optimal candidate assignment sequence obtained by consecutively setting the number of iterations is consistent. Stop iterating and determine the optimal candidate assignment sequence obtained in the last iteration as the final candidate assignment sequence. The final candidate dispatch sequence is filled into the position corresponding to the machine's non-locked state, while keeping the reserved section position corresponding to the machine's locked state unchanged, thus generating the disturbance-corrected dispatch sequence for the corresponding PCB drilling machine.

[0014] Optionally, the handover generation module specifically comprises: Based on the disturbance correction dispatch sequence of each PCB drilling machine, the task execution order corresponding to each PCB drilling machine is extracted. After each PCB drilling machine completes the adaptation execution section corresponding to the current processing segment, the next adaptation execution section after the current processing segment is selected from the disturbance correction dispatch sequence of the corresponding PCB drilling machine as the next stage task. According to the task execution order of the disturbance correction dispatch sequence, the time connection of the adaptation execution segments of each PCB drilling machine is sequentially performed to form a continuous processing sequence of each PCB drilling machine. The continuous processing sequences of each PCB drilling machine are summarized to generate a continuous processing handover plan.

[0015] The beneficial effects of this invention are: This invention addresses the problems of insufficient utilization of process features, lagging equipment status response, and difficulty in dynamically optimizing scheduling results in the existing PCB drilling machine dispatching process by improving the synergistic integration of the Resonance model and the penetration-driven Jaya scheduling algorithm. In the process deconstruction module, PCB drilling tasks are structured and a PCB drilling task process set is generated, making the task feature representation more complete. In the capability building module, a real-time capability set of the drilling machine is constructed, enabling real-time perception of equipment operating status and its participation in subsequent decision-making. In the segment division module, a task tool execution segment table is formed, ensuring the continuity of processing segments and reducing path disturbances. In the adaptation calculation module, the task process spectral sequence and equipment response spectral sequence are constructed by improving the Resonance model, and a harmonic window penetration sequence and normalized contribution value are generated by combining a temporal harmonic window penetration mechanism, ensuring that the task-equipment adaptation list is accurate. It accurately reflects the overall adaptability of the target execution segment; in the sequence arrangement module, a drilling machine sequence dispatch table is generated to improve the continuity of task organization; in the position locking module, a machine position dispatch instruction set is generated to give the processing process a stable time constraint structure; further, in the disturbance optimization module, a local rearrangement segment sequence is constructed based on the machine position locking state, and iterative optimization is performed driven by the sequence evaluation value through the penetration-driven Jaya scheduling algorithm, so that the optimization process is simultaneously constrained by the normalized contribution value and the state penalty value, realizing the joint adjustment of the equipment operating state and the degree of task adaptability, and finally generating the disturbance correction dispatch sequence for the corresponding PCB drilling machine; in the handover generation module, a continuous processing handover plan is formed based on the disturbance correction dispatch sequence, so that the task acceptance process of each PCB drilling machine is continuous and smooth, thereby reducing the risk of equipment vibration fluctuation and abnormal tool wear, and improving processing stability and production efficiency. Attached Figure Description

[0016] 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 This is a schematic diagram of the intelligent dispatch management system for PCB drilling machines based on machine learning proposed in this invention. Figure 2 This is a schematic diagram of the improved Resonance model structure of the intelligent dispatch management system for PCB drilling machines based on machine learning proposed in this invention. Detailed Implementation

[0017] 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.

[0018] refer to Figure 1 and Figure 2 The machine learning-based intelligent dispatch management system for PCB drilling machines includes the following modules: The process deconstruction module is used to receive PCB orders to be produced and corresponding drilling process files, deconstruct the process of each PCB drilling task, and generate a PCB drilling task process set. The capability building module is used to collect the operating status data of each PCB drilling machine and build a real-time capability set for the drilling machine. The segment division module is used to divide the tool execution segments of each PCB drilling task on each PCB drilling machine based on the PCB drilling task process set and the drilling machine real-time capability set, forming a task tool execution segment table. The adaptation calculation module is used to input the task tool execution segment table and the drilling machine real-time capability set into the improved Resonance model for processing to obtain the task-equipment adaptation list. The improved Resonance model includes a process spectrum generation unit, an equipment response spectrum generation unit, a harmonic window penetration unit, and a dispatch adaptation output unit. The harmonic window penetration unit has a built-in time-series harmonic window penetration mechanism. The sequence arrangement module is used to arrange the PCB drilling tasks in sequence based on the task-equipment adaptation list and generate a drilling machine sequence dispatch table. The occupancy locking module is used to perform machine occupancy locking operations on each PCB drilling machine based on the drilling machine sequence dispatch table, and generate a set of machine occupancy dispatch instructions. The disturbance optimization module is used to update the operating status data of each PCB drilling machine during the execution of the machine occupancy dispatch instruction set, and perform local rearrangement based on the penetration-driven Jaya scheduling algorithm to generate the disturbance correction dispatch sequence for the corresponding PCB drilling machine. The handover generation module is used to take over the next stage of tasks for PCB drilling machines that have completed the current processing segment based on the disturbance correction dispatch sequence, and generate a continuous processing handover plan.

[0019] In this embodiment, the process deconstruction module specifically comprises: Receive PCB orders to be produced and corresponding drilling process documents, parse the drilling process documents, and extract hole position coordinate data, hole group distribution area data, board thickness segment data, hole diameter data, and hole sequence path data; Based on the aperture size data, aperture levels are divided to obtain aperture level data; The drilling path length is calculated based on the hole position coordinate data and hole sequence path data. The drilling path length is the cumulative distance between adjacent hole position coordinates. The number of path switching times is calculated based on the hole sequence path data, where the number of path switching times is the number of times adjacent processing directions change in the hole sequence path; The hole diameter level data, hole group distribution area data, board thickness section data, drilling path length and path switching number are structured and combined to generate a PCB drilling task process set. In the specific implementation process, the drilling process document is preferably an Excellon format file. The hole position coordinate data and hole sequence path data are obtained by parsing line by line. The hole position coordinate data is represented by two-dimensional coordinates in millimeters. The hole diameter data is divided into three hole diameter levels according to the diameter range: micro hole level with a diameter of less than or equal to 0.2 mm, medium hole level with a diameter of greater than 0.2 mm and less than or equal to 0.6 mm, and large hole level with a diameter of greater than 0.6 mm. The drilling path length is obtained by accumulating the straight-line distance between adjacent holes segment by segment. The number of path switching is counted by judging whether the included angle of the processing direction formed by three adjacent points is greater than 30 degrees.

[0020] In this embodiment, the operating status data includes spindle speed data, spindle vibration amplitude data, spindle temperature data, current tool diameter data, and machine tool occupancy status data; In the specific implementation process, spindle speed data is collected in real time by an encoder installed at the end of the spindle, with a sampling frequency set to 1000 times per second, to reflect the speed stability of the drilling machine at different processing stages; spindle vibration amplitude data is obtained by an accelerometer installed on the spindle support structure, and the acceleration signal is converted into vibration amplitude, with a sampling frequency set to 2000 times per second, to characterize the vibration state of the equipment during drilling; spindle temperature data is collected by a temperature sensor arranged on the spindle housing, with a sampling period set to once every 1 second, to reflect thermal drift; current tool diameter data is directly read from the tool specification information in the tool management system; machine occupancy status data is obtained through the equipment control system and identified as either occupied or idle; by aligning the above operating status data with a unified timestamp and storing it in a structured manner, the constructed real-time capability set of the drilling machine can truly reflect the current processing capacity of the equipment, providing a stable data foundation for task allocation, thereby reducing unreasonable work assignments caused by equipment status fluctuations.

[0021] In this embodiment, the segmentation module specifically comprises: Based on the hole diameter level data, hole group distribution area data, board thickness segment data, drilling path length and path switching number in the PCB drilling task process set, each PCB drilling task is segmented according to the hole diameter level change position and hole sequence path order to obtain the initial processing segment set. Based on the current tool diameter data in the real-time capability set of the drilling machine, the hole diameter is matched for each processing segment in the initial processing segment set. The processing segment whose hole diameter level data is consistent with the current tool diameter data is mapped to the corresponding PCB drilling machine to obtain the first execution segment set. Based on the hole group distribution area data, each processing segment in the first execution segment set is grouped according to its hole group distribution area, and processing segments that are consecutive in the same hole group distribution area and in the hole sequence path are merged to obtain the second execution segment set. Based on the plate thickness segment data, each processing segment in the second execution segment set is divided according to the plate thickness segment, and processing segments with the same plate thickness segment are merged to obtain the third execution segment set; Based on the drilling path length and the number of path switching, the processing segments in the third execution segment set are rearranged according to the path order, and the processing segments with adjacent paths that do not increase the number of path switching after splicing are spliced ​​together to obtain the fourth execution segment set. Based on the spindle speed data, spindle vibration amplitude data, and spindle temperature data, the operating status of each processing segment in the fourth execution segment set is matched, and the processing segment corresponding to the operating status is marked as the matching segment to obtain the fifth execution segment set; Based on the machine occupancy status data, the occupancy status of each processing segment in the fifth execution segment set is determined, and the processing segments with idle machine occupancy status are retained to obtain the target execution segment set; The target execution segment set is combined according to the correspondence between PCB drilling tasks and PCB drilling machines to form a task tool execution segment table; In the specific implementation process, the initial processing segment set is generated point by point according to the hole sequence path. When the hole diameter level of adjacent holes changes or the path direction changes by more than 30 degrees, a new processing segment is defined. During the hole diameter matching process, the current tool diameter is compared one by one with the corresponding range of hole diameter levels, and the matching deviation is controlled within 0.01 mm. When grouping the hole group distribution area, it is classified according to the area number defined in the process document, and only the processing segments with consecutive numbers in the same area are merged. When processing plate thickness segments, processing segments with a plate thickness difference of less than or equal to 0.05 mm are merged into the same group to ensure processing accuracy. To improve stability, during path rearrangement, the holes are arranged sequentially based on the hole sequence path, and splicing is only performed on adjacent sections with a spatial distance of less than or equal to 2 mm and a directional change of less than 30 degrees to avoid additional path disturbances. During operation matching, machining sections with spindle speed fluctuations controlled within ±5%, vibration amplitude fluctuations controlled within ±0.005 mm, and temperature fluctuations controlled within ±2 degrees Celsius are marked as matching sections. Through the above processing, the resulting task tool execution section table has the characteristics of strong continuity, small path disturbances, and stable equipment operation, thereby reducing tool change frequency and improving drilling accuracy.

[0022] In this embodiment, the adaptation calculation module specifically comprises: The task tool execution segment table is input into the process spectrum generation unit. The target execution segment set corresponding to each PCB drilling task and PCB drilling machine in the task tool execution segment table is expanded. The hole diameter level data, hole group distribution area data, board thickness segment data, drilling path length and path switching number in each target execution segment are arranged in order to form a multi-dimensional process vector. The value of each dimension is divided by the maximum value of the corresponding dimension to generate the task process spectrum and construct the task process spectrum sequence. The PCB drilling machine correspondence in the task tool execution segment table is input into the device response spectrum generation unit. Based on the spindle speed data, spindle vibration amplitude data, spindle temperature data, current tool diameter data, and machine occupancy status data in the real-time capability set of the drilling machine, the operating status data of the PCB drilling machine corresponding to the target execution segment is extracted one by one. The operating status data are arranged into a multi-dimensional response vector in chronological order. The value of each dimension is divided by the maximum value of the corresponding dimension to generate the device response spectrum and construct the device response spectrum sequence. In the specific implementation process, the multidimensional process vector is constructed using a fixed-length vector encoding method. The pore size level data is encoded using a discrete numbering method, with micropore level encoded as 1, mesopore level encoded as 2, and macropore level encoded as 3. The pore group distribution area data is directly mapped to integer encoding according to the area number. The multidimensional response vector is represented using a fixed-length vector of the same dimension. The machine occupancy status data is encoded by marking occupancy as 1 and idle as 0, so that the process characteristics and equipment response characteristics can be matched and calculated in the same numerical space, improving the stability and consistency of spectral element calculation. The task process spectrum sequence and the equipment response spectrum sequence are input into the harmonic window penetration unit. Each process spectrum in the task process spectrum sequence is subtracted from the corresponding PCB drilling machine's equipment response spectrum to generate a spectrum difference matrix. The values ​​less than zero in the spectrum difference matrix are set to zero to generate a non-negative response matrix. The timing harmonic window penetration mechanism is called on the non-negative response matrix for processing to generate a harmonic window penetration sequence. The harmonic window penetration sequence is input into the dispatch adapter output unit, and the harmonic window penetration sequence is subjected to forward accumulation operation according to the target execution segment order to generate the accumulated response value of the corresponding target execution segment. The cumulative response values ​​corresponding to all target execution segments are summed to obtain the total response. The cumulative response value corresponding to each target execution segment is divided by the total response to generate the normalized contribution value of the corresponding target execution segment. Select the target execution segment whose normalized contribution value is greater than or equal to the set contribution threshold as the adapted execution segment, and generate a task-device adaptation list; The improved Resonance model in this invention is a structured extension of the traditional Resonance model. The traditional Resonance model mainly calculates the response based on the matching relationship between input features. Its input is usually a single-dimensional or low-dimensional feature vector, focusing on the similarity analysis between static features. However, this invention introduces a task tool execution segment table as the input carrier, directly embedding the correspondence between PCB drilling tasks and PCB drilling machines into the model input structure. Through the process spectrum generation unit, the hole diameter level data, hole group distribution area data, board thickness segment data, drilling path length, and path switching times are uniformly expanded into multi-dimensional process vectors. Through the equipment response spectrum generation unit, the spindle speed data, spindle vibration amplitude data, spindle temperature data, current tool diameter data, and machine occupancy status data are uniformly expanded into multi-dimensional response vectors. Thus, a dual-sequence structure of task process spectrum sequence and equipment response spectrum sequence is constructed, realizing the structural transformation from single feature matching to dual-dimensional response matching of process features and equipment status.

[0023] Based on this, by performing dimension-wise difference calculations on the task process spectrum sequence and the equipment response spectrum sequence and constructing a non-negative response matrix, the model can suppress mismatch features. Simultaneously, by performing forward accumulation on the harmonic window penetration sequence and further calculating the normalized contribution value, the overall fit of each target execution segment is quantitatively expressed, avoiding bias caused by relying solely on local features for judgment. By setting a uniform contribution threshold of 0.3, the output task-equipment fit list has clear filtering boundaries, thereby improving the stability and consistency of the dispatch results. Compared with the traditional Resonance model, this improved scheme establishes a complete processing flow in terms of input structure, response expression, and output decision-making, encompassing process data processing, equipment state matching, and fit result generation. This makes the allocation of PCB drilling tasks among different drilling machines more consistent with the actual operating status of the equipment and the requirements of processing continuity, thereby reducing equipment vibration, abnormal tool wear, and decreased processing efficiency caused by mismatch allocation.

[0024] In this embodiment, the timing harmonic window pass-through mechanism is specifically as follows: For each row of data in the non-negative response matrix, perform a dimension-wise subtraction operation with the adjacent previous row of data. When the current row is the first row, use the current row of data as the adjacent previous row of data for the operation to generate the forward original difference matrix. Set the values ​​less than zero in the forward original difference matrix to zero to generate the forward difference matrix. For each row of data in the non-negative response matrix, perform a dimension-wise subtraction operation with the adjacent next row of data. When the current row is the last row, treat the current row of data as the adjacent next row of data for the operation to generate the backward original difference matrix. Set the values ​​less than zero in the backward original difference matrix to zero to generate the backward difference matrix. The forward difference matrix and the backward difference matrix are multiplied to generate a bidirectional coupling matrix; Each element in the bidirectional coupling matrix is ​​incremented by 1 and subjected to a natural logarithm operation, wherein the natural logarithm operation is a logarithmic operation with the constant e as the base and the corresponding element value as the argument, thereby generating a logarithmic coupling matrix; Multiply each element in the logarithmic coupling matrix by the reciprocal of the corresponding target execution segment index value to generate a scaled logarithmic matrix; An exponential function operation is performed on each element in the scaled logarithmic matrix. The exponential function operation is a power operation with a constant e as the base and the corresponding element value as the exponent, to generate an exponential response matrix. Normalize the data in each row of the exponential response matrix by dividing each element in each row by the sum of all elements in the corresponding row to generate a normalized response matrix. Each row of data in the normalized response matrix is ​​multiplied dimension-by-dimensionally with the reciprocal of the corresponding target execution segment index value to generate a position-weighted response matrix; Perform a dimension-by-dimensional summation operation on each row of data in the position-weighted response matrix to generate a penetration response value, and arrange them according to the target execution segment order to generate a harmonic window penetration sequence; This invention introduces a time-series harmonic window penetration mechanism to expand the difference between the process response and equipment response in the non-negative response matrix in both forward and backward directions along the time dimension. Through dimension-wise subtraction, non-negative truncation, and bidirectional coupling operations, it effectively extracts the stable variation characteristics of each execution segment during continuous processing. Furthermore, by constructing a nonlinear transformation process using logarithmic and exponential function operations, it compresses the influence of extreme values ​​while maintaining the magnitude regularity of response data of different amplitudes, thereby improving the ability to identify medium-response segments. Combined with scaling processing of the reciprocal of the execution segment index value and bitwise operations… Weighted operations are applied to give higher weight to the execution segments of the preceding targets in the overall calculation, strengthening the scheduling strategy that prioritizes processing continuity. By accumulating the normalized response matrix dimension by dimension to generate penetration response values, a unified quantitative expression of the overall adaptability of each target execution segment is achieved, avoiding misjudgments caused by a single moment or a single feature. The resulting harmonic window penetration sequence can accurately reflect the continuous adaptability of each execution segment in the actual processing process, thereby improving the stability of task allocation, reducing the risk of vibration fluctuations and abnormal tool wear during drilling, and improving the overall processing efficiency and consistency of processing quality.

[0025] In this embodiment, the sequential arrangement module specifically comprises: Based on the task-equipment adaptation list, the adaptation execution segments corresponding to each PCB drilling task are grouped according to the PCB drilling machine to generate a machine task set. In the set of machine tasks, the adaptation execution segments are grouped according to the aperture level data. Adaptive execution segments with the same aperture level data are grouped into the same group to generate an aperture group execution sequence. In the aperture grouping execution sequence, the adaptation execution segments are grouped according to the aperture distribution area data, and the adaptation execution segments with the same aperture distribution area data are divided into the same group to generate a region grouping execution sequence; In the aforementioned regional grouping execution sequence, the adaptation execution segments are grouped according to the plate thickness segment data. Adaptive execution segments with the same plate thickness segment data are grouped into the same group to generate a plate thickness grouping execution sequence. The adaptation execution segments within each group of the plate thickness grouping execution sequence are sorted from largest to smallest according to their corresponding normalized contribution values ​​to generate a sorted execution sequence within the group. The maximum normalized contribution value within each group is used as the inter-group ranking value, and the groups are sorted from largest to smallest according to the inter-group ranking value to generate an inter-group ranking sequence. Arrange the intra-group sorting execution sequences of each group according to the inter-group sorting sequence to form the task execution order of the corresponding PCB drilling machine. The task execution order of each PCB drilling machine is summarized to form a drilling machine sequence dispatch table. This invention organizes the adaptation execution segments into a multi-level structure before dispatching, ensuring that task allocation is not only based on a single degree of adaptation but also considers key process factors such as hole diameter level, hole group distribution area, and board thickness segment. This reduces frequent switching between different processing conditions during scheduling. Normalized contribution value sorting is introduced within each group, prioritizing execution segments with higher adaptation degrees within each group, improving the stability of the single machine's processing. Furthermore, by using the maximum normalized contribution value within a group as the basis for inter-group sorting, groups including high-value execution segments are given priority in entering the processing sequence, thus improving the overall processing priority of critical tasks. Through the synergistic effect of the above multi-level sorting and grouping, the generated drilling machine sequential dispatch table has stronger processing continuity and task concentration, effectively reducing tool changes and machine switching frequency, minimizing vibration fluctuations during processing, and improving equipment utilization efficiency and the consistency of PCB drilling processing quality.

[0026] In this embodiment, the occupant locking module specifically comprises: Based on the drilling machine sequence dispatch table, the task execution order corresponding to each PCB drilling machine is extracted, and the corresponding appropriate execution segment is determined one by one according to the task execution order. For each PCB drilling machine, the adaptation execution segment is allocated a continuous time period according to the task execution order to generate a machine occupancy time sequence. Bind each time period in the machine occupancy time sequence to the corresponding adaptive execution segment to generate a machine occupancy mapping relationship; Based on the machine occupancy mapping relationship, a machine occupancy locking operation is performed on each PCB drilling machine. The PCB drilling machine corresponding to the time period within the machine occupancy time sequence is marked as locked; the PCB drilling machine corresponding to the time period not within the machine occupancy time sequence is marked as unlocked. The machine occupancy mapping relationship and machine occupancy locking status are combined to generate a machine occupancy dispatch instruction set; This invention introduces a machine position locking operation based on the drilling machine sequential dispatch table, transforming the task execution order from a simple sorting result into an executable instruction structure with time constraints. By binding the appropriate execution segment with continuous time periods, a clear machine position time sequence is formed, making the processing arrangement of each PCB drilling machine predictable within the future time range. Furthermore, by locking the time periods within the machine position time sequence, the corresponding machine cannot be inserted by other tasks during the locked time, effectively avoiding processing order disturbances caused by temporary task adjustments. At the same time, by distinguishing between locked and unlocked states, the scheduling system can flexibly arrange new tasks during unlocked time periods, thus ensuring the continuity of critical tasks while taking into account scheduling flexibility. The final generated machine position dispatch instruction set can achieve stable execution of task arrangements, reduce machine switching frequency and scheduling conflicts, and improve the continuity of the PCB drilling process and overall production efficiency.

[0027] In this embodiment, the disturbance optimization module specifically comprises: Real-time updates of the operating status data of each PCB drilling machine, generating and updating the real-time capability set of the drilling machine; Based on the machine position locking status in the machine position dispatch instruction set, the corresponding adaptation execution section is marked as a reserved section, and the corresponding adaptation execution section is marked as a rearrangement section. According to the task execution order in the drilling machine sequence dispatch table, extract the rearrangement segment corresponding to each PCB drilling machine and arrange them according to the task execution order to generate a local rearrangement segment sequence. For each PCB drilling machine, the penetration-driven Jaya scheduling algorithm is executed. Each rearranged segment in the local rearranged segment sequence is sequentially numbered according to the task execution order, and the sequential number is used as the position number. Each rearranged segment is assigned a unique identifier. For each pair of adjacent rearranged segments in the local rearranged segment sequence, an adjacent swap operation is performed to generate multiple initial candidate dispatch sequences. The adjacent swap operation is to swap the positions of the nth rearranged segment and the (n+1)th rearranged segment in the local rearranged segment sequence, while keeping the positions of the remaining rearranged segments unchanged. The sum of contribution values ​​is obtained by summing the normalized contribution values ​​corresponding to each rearranged segment in the initial candidate dispatch sequence; the state penalty value is obtained by adding the normalized value of the spindle vibration amplitude data of the PCB drilling machine corresponding to the updated real-time capability set of the drilling machine to the normalized value of the spindle temperature data. The normalized value of the spindle vibration amplitude data is obtained by dividing the current spindle vibration amplitude data of the PCB drilling machine by the maximum spindle vibration amplitude collected during the operation cycle of the PCB drilling machine; the normalized value of the spindle temperature data is obtained by dividing the current spindle temperature data of the PCB drilling machine by the maximum spindle temperature collected during the operation cycle of the PCB drilling machine; the above two normalized values ​​are added together to obtain the state penalty value, which is used to characterize the operating load level of the current PCB drilling machine; Subtract the state penalty value from the sum of the contribution values ​​to obtain the sequence evaluation value. The initial candidate assignment sequence with the largest sequence evaluation value is determined as the optimal candidate assignment sequence, and the initial candidate assignment sequence with the smallest sequence evaluation value is determined as the worst candidate assignment sequence. Based on the unique identifier of each rearranged segment, the position number of the corresponding rearranged segment is found in both the best and worst candidate assignment sequences. The difference between the current position number and the corresponding position number in the best candidate assignment sequence is calculated and its absolute value is taken to obtain the first position difference. The difference between the current position number and the corresponding position number in the worst candidate assignment sequence is calculated and its absolute value is taken to obtain the second position difference. The current position number is added to the first position difference and the second position difference is subtracted to obtain the updated position number. If the updated position number is outside the valid range, a modulo operation is performed. The rearranged segments are reordered according to the updated position number to generate an updated candidate assignment sequence, and the updated candidate assignment sequence is used as the initial candidate assignment sequence for the next iteration. Repeatedly perform sequence evaluation value calculation, optimal candidate assignment sequence determination, worst candidate assignment sequence determination, and update position number calculation until the maximum number of iterations or the optimal candidate assignment sequence obtained by consecutively setting the number of iterations is consistent, then stop iterating, and determine the optimal candidate assignment sequence obtained by the last iteration as the final candidate assignment sequence; preferably, the maximum number of iterations is 10, and the set number of iterations is 3. The final candidate dispatch sequence is filled into the position corresponding to the machine's non-locked state, while keeping the reserved section position corresponding to the machine's locked state unchanged, thus generating the disturbance correction dispatch sequence for the corresponding PCB drilling machine. This invention introduces a real-time updated set of drilling machine capabilities during the execution of the machine positioning and dispatching instruction set. This allows the scheduling process to dynamically perceive changes in the operating status of each PCB drilling machine, converting spindle vibration amplitude data and spindle temperature data into state penalty values ​​for sequence evaluation. This ensures that dispatching optimization considers not only the process suitability of the task itself but also the current operating load of the equipment, avoiding the continued allocation of high-intensity processing tasks under high load conditions. By distinguishing between machine positioning locked and machine positioning unlocked states, the original order of key execution sections is maintained without adjustment, and only locally optimized rearranged sections are performed, ensuring processing continuity and improving scheduling flexibility. Furthermore, multiple initial candidate dispatching sequences are generated through adjacent exchanges, and the candidate sequences are continuously updated during the iteration process, allowing the optimization process to gradually approach a better solution within a limited search space, improving the stability and feasibility of the dispatching results.

[0028] The improvement of the penetration-driven Jaya scheduling algorithm lies in extending the traditional Jaya algorithm, which is based solely on solution space position updates, to a composite evaluation-driven mechanism that combines process contribution and equipment status. By introducing normalized contribution values ​​and state penalty values ​​to construct sequence evaluation values, the update direction is simultaneously constrained by both task suitability and equipment operating status. Furthermore, a unique identifier is used to establish the correspondence between rearranged segments and different candidate dispatch sequences, achieving accurate updates based on position differences and avoiding disordered disturbances. Penetration-driven refers to the fact that sequence evaluation values ​​not only affect the overall sequence quality judgment but also participate in position update calculations, allowing evaluation information to penetrate from the global level to the position adjustment process of individual rearranged segments. This forms a scheduling method where evaluation results drive structural optimization, thereby improving optimization efficiency, reducing ineffective iterations, and ultimately improving the processing continuity, equipment stability, and overall efficiency of the generated disturbance-corrected dispatch sequence.

[0029] In this embodiment, the handover generation module specifically comprises: Based on the disturbance correction dispatch sequence of each PCB drilling machine, the task execution order corresponding to each PCB drilling machine is extracted. After each PCB drilling machine completes the adaptation execution section corresponding to the current processing segment, the next adaptation execution section after the current processing segment is selected from the disturbance correction dispatch sequence of the corresponding PCB drilling machine as the next stage task. According to the task execution order of the disturbance correction dispatch sequence, the time connection of the adaptation execution segments of each PCB drilling machine is sequentially performed to form a continuous processing sequence of each PCB drilling machine. The continuous processing sequences of each PCB drilling machine are summarized to generate a continuous processing handover plan.

[0030] Example 1: To verify the feasibility of this invention in practice, it was applied to a drilling dispatching scenario in a PCB manufacturing workshop. This workshop is equipped with 12 six-axis PCB drilling machines, processing approximately 460 batches of high-density interconnect boards, communication backplanes, and ordinary multilayer boards daily. The original dispatching method relied mainly on manual experience and simple order priority rules, often resulting in problems such as tasks with the same hole diameter being distributed across multiple machines, tasks with significant differences in board thickness being processed continuously, and high-load tasks being assigned even after the equipment temperature had risen. This led to numerous tool changes, long machine waiting times, and significant fluctuations in hole position deviation.

[0031] In this embodiment, the system first receives the PCB order and Excellon drilling process file from the ERP system, parses them to obtain the PCB drilling task process set, and then collects the operating status data to form the real-time capability set of the drilling machine. Subsequently, the segmentation module generates a task tool execution segment table based on the hole diameter level, hole group area, board thickness segment, and path change. The adaptation calculation module inputs this table into the improved Resonance model, and through the process spectrum generation unit, the equipment response spectrum generation unit, and the harmonic window penetration unit with the built-in timing harmonic window penetration mechanism, the task is obtained. - Equipment compatibility list; During work assignment, the system filters suitable execution segments based on normalized contribution values, setting a contribution threshold of 0.3 to effectively filter invalid schedules with low compatibility, reducing the probability of secondary adjustments. The sequential arrangement module generates a drilling machine sequential work assignment table. During execution, the disturbance optimization module uses the penetration-driven Jaya scheduling algorithm to adjust only the non-locked segments, keeping the locked segments unchanged, generating the disturbance-corrected work assignment sequence for the corresponding PCB drilling machines, and taking over the next stage of tasks for the PCB drilling machines that have completed the current processing segment, generating a continuous processing handover plan. In actual operation, the penetration-driven Jaya algorithm can reach a convergence state with consistent optimal solutions for three consecutive times within an average of 5 iterations, ensuring the real-time performance of dynamic rescheduling.

[0032] To verify the actual effect of the present invention, three comparison schemes were set up. Comparison scheme one is a manual rule-based work assignment scheme, which assigns work only based on the order delivery date and aperture. Comparison scheme two is a traditional Resonance model work assignment scheme, which only uses static task and equipment matching. Comparison scheme three is a traditional Resonance model combined with the ordinary Jaya algorithm scheme. The system of the present invention adopts an improved Resonance model and a penetration-driven Jaya scheduling algorithm. The comparison results are shown in Table 1.

[0033] Table 1. Comparison of PCB drilling production performance under different dispatch schemes As can be seen from the data in Table 1 above, the system of the present invention outperforms comparative schemes 1 to 3 in all key indicators, demonstrating a significant improvement in overall performance. In terms of dispatching calculation efficiency, the average dispatching calculation time of the system of this invention is 9.5s, which is about 48.9% shorter than the 18.6s of the first comparative scheme. This shows that the improved Resonance model and the penetration-driven Jaya scheduling algorithm can effectively reduce the computational complexity and improve the response speed. In terms of processing continuity, the average number of tool changes is reduced from 42 times in the first comparative scheme to 23 times, a reduction of 45.2%. At the same time, the machine waiting time is reduced from 73.5min to 31.6min. This shows that the synergistic effect of the task tool execution segment table and the sequence arrangement module significantly reduces invalid switching and waiting time. In terms of processing quality, the average hole position deviation is reduced from 18.4μm to 10.8μm, and the abnormal tool wear rate is reduced from 6.8% to 2.7%. This shows that after introducing the drilling machine real-time capability set and state penalty mechanism, the equipment operating status is effectively constrained, reducing the impact of vibration and thermal drift. In terms of production capacity, the system of this invention completes an average of 462 batches per day, which is about 7.2% higher than the third comparative scheme, further demonstrating the overall scheduling optimization effect.

[0034] As can be seen from this embodiment, the present invention achieves deep integration of process characteristics and equipment operating status during the PCB drilling machine dispatching process. By constructing a PCB drilling task process set and a drilling machine real-time capability set, the task division and equipment matching are made more refined. By leveraging the improved Resonance model and its built-in time-series harmonic window penetration mechanism, the accuracy and continuity of the task-equipment adaptation list are improved. At the same time, by combining machine station occupancy locking operation and penetration-driven Jaya scheduling algorithm, the dispatching sequence is dynamically optimized, improving scheduling flexibility while ensuring stable execution of key processing sections. Finally, by perturbation correction of the dispatching sequence and continuous processing handover plan, smooth task connection between PCB drilling machines is achieved, thereby improving overall processing continuity, equipment utilization efficiency, and processing quality consistency.

[0035] 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 machine learning-based intelligent dispatch management system for PCB drilling machines, characterized in that, Includes the following modules: The process deconstruction module is used to receive PCB orders to be produced and corresponding drilling process files, deconstruct the process of each PCB drilling task, and generate a PCB drilling task process set. The capability building module is used to collect the operating status data of each PCB drilling machine and build a real-time capability set for the drilling machine. The segment division module is used to divide the tool execution segments of each PCB drilling task on each PCB drilling machine based on the PCB drilling task process set and the drilling machine real-time capability set, forming a task tool execution segment table. The adaptation calculation module is used to input the task tool execution segment table and the drilling machine real-time capability set into the improved Resonance model for processing to obtain the task-equipment adaptation list. The improved Resonance model includes a process spectrum generation unit, an equipment response spectrum generation unit, a harmonic window penetration unit, and a dispatch adaptation output unit. The harmonic window penetration unit has a built-in time-series harmonic window penetration mechanism. The sequence arrangement module is used to arrange the PCB drilling tasks in sequence based on the task-equipment adaptation list and generate a drilling machine sequence dispatch table. The occupancy locking module is used to perform machine occupancy locking operations on each PCB drilling machine based on the drilling machine sequence dispatch table, and generate a set of machine occupancy dispatch instructions. The disturbance optimization module is used to update the operating status data of each PCB drilling machine during the execution of the machine occupancy dispatch instruction set, and perform local rearrangement based on the penetration-driven Jaya scheduling algorithm to generate the disturbance correction dispatch sequence for the corresponding PCB drilling machine. The handover generation module is used to take over the next stage of tasks for PCB drilling machines that have completed the current processing segment based on the disturbance correction dispatch sequence, and generate a continuous processing handover plan.

2. The intelligent dispatch management system for PCB drilling machines based on machine learning as described in claim 1, characterized in that, The process deconstruction module is specifically as follows: Receive PCB orders to be produced and corresponding drilling process documents, parse the drilling process documents, and extract hole position coordinate data, hole group distribution area data, board thickness segment data, hole diameter data, and hole sequence path data; Based on the aperture size data, aperture levels are divided to obtain aperture level data; The drilling path length is calculated based on the hole position coordinate data and hole sequence path data. The drilling path length is the cumulative distance between adjacent hole position coordinates. The number of path switching times is calculated based on the hole sequence path data, where the number of path switching times is the number of times adjacent processing directions change in the hole sequence path; The hole diameter level data, hole group distribution area data, board thickness section data, drilling path length, and path switching number are structured and combined to generate a PCB drilling task process set.

3. The intelligent dispatch management system for PCB drilling machines based on machine learning as described in claim 1, characterized in that, The operating status data includes spindle speed data, spindle vibration amplitude data, spindle temperature data, current tool diameter data, and machine tool occupancy status data.

4. The intelligent dispatch management system for PCB drilling machines based on machine learning according to claim 3, characterized in that, The segmentation module is specifically as follows: Based on the hole diameter level data, hole group distribution area data, board thickness segment data, drilling path length and path switching number in the PCB drilling task process set, each PCB drilling task is segmented according to the hole diameter level change position and hole sequence path order to obtain the initial processing segment set. Based on the current tool diameter data in the real-time capability set of the drilling machine, the hole diameter is matched for each processing segment in the initial processing segment set. The processing segment whose hole diameter level data is consistent with the current tool diameter data is mapped to the corresponding PCB drilling machine to obtain the first execution segment set. Based on the hole group distribution area data, each processing segment in the first execution segment set is grouped according to its hole group distribution area, and processing segments that are consecutive in the same hole group distribution area and in the hole sequence path are merged to obtain the second execution segment set. Based on the plate thickness segment data, each processing segment in the second execution segment set is divided according to the plate thickness segment, and processing segments with the same plate thickness segment are merged to obtain the third execution segment set; Based on the drilling path length and the number of path switching, the processing segments in the third execution segment set are rearranged according to the path order, and the processing segments with adjacent paths that do not increase the number of path switching after splicing are spliced ​​together to obtain the fourth execution segment set. Based on the spindle speed data, spindle vibration amplitude data, and spindle temperature data, the operating status of each processing segment in the fourth execution segment set is matched, and the processing segment corresponding to the operating status is marked as the matching segment to obtain the fifth execution segment set; Based on the machine occupancy status data, the occupancy status of each processing segment in the fifth execution segment set is determined, and the processing segments with idle machine occupancy status are retained to obtain the target execution segment set; The target execution segment set is combined according to the correspondence between PCB drilling tasks and PCB drilling machines to form a task tool execution segment table.

5. The intelligent dispatch management system for PCB drilling machines based on machine learning according to claim 1, characterized in that, The adaptation calculation module is specifically as follows: The task tool execution segment table is input into the process spectrum generation unit. The target execution segment set corresponding to each PCB drilling task and PCB drilling machine in the task tool execution segment table is expanded. The hole diameter level data, hole group distribution area data, board thickness segment data, drilling path length and path switching number in each target execution segment are arranged in order to form a multi-dimensional process vector. The value of each dimension is divided by the maximum value of the corresponding dimension to generate the task process spectrum and construct the task process spectrum sequence. The PCB drilling machine correspondence in the task tool execution segment table is input into the device response spectrum generation unit. Based on the spindle speed data, spindle vibration amplitude data, spindle temperature data, current tool diameter data, and machine occupancy status data in the real-time capability set of the drilling machine, the operating status data of the PCB drilling machine corresponding to the target execution segment is extracted one by one. The operating status data are arranged into a multi-dimensional response vector in chronological order. The value of each dimension is divided by the maximum value of the corresponding dimension to generate the device response spectrum and construct the device response spectrum sequence. The task process spectrum sequence and the equipment response spectrum sequence are input into the harmonic window penetration unit. Each process spectrum in the task process spectrum sequence is subtracted from the corresponding PCB drilling machine's equipment response spectrum to generate a spectrum difference matrix. The values ​​less than zero in the spectrum difference matrix are set to zero to generate a non-negative response matrix. The timing harmonic window penetration mechanism is called on the non-negative response matrix for processing to generate a harmonic window penetration sequence. The harmonic window penetration sequence is input into the dispatch adapter output unit, and the harmonic window penetration sequence is subjected to forward accumulation operation according to the target execution segment order to generate the accumulated response value of the corresponding target execution segment. The cumulative response values ​​corresponding to all target execution segments are summed to obtain the total response. The cumulative response value corresponding to each target execution segment is divided by the total response to generate the normalized contribution value of the corresponding target execution segment. Select the target execution segment whose normalized contribution value is greater than or equal to the set contribution threshold as the adapted execution segment, and generate the task-device adaptation list.

6. The intelligent dispatch management system for PCB drilling machines based on machine learning according to claim 5, characterized in that, The specific timing harmonic window penetration mechanism is as follows: For each row of data in the non-negative response matrix, perform a dimension-wise subtraction operation with the adjacent previous row of data. When the current row is the first row, use the current row of data as the adjacent previous row of data for the operation to generate the forward original difference matrix. Set the values ​​less than zero in the forward original difference matrix to zero to generate the forward difference matrix. For each row of data in the non-negative response matrix, perform a dimension-wise subtraction operation with the adjacent next row of data. When the current row is the last row, treat the current row of data as the adjacent next row of data for the operation to generate the backward original difference matrix. Set the values ​​less than zero in the backward original difference matrix to zero to generate the backward difference matrix. The forward difference matrix and the backward difference matrix are multiplied to generate a bidirectional coupling matrix; Each element in the bidirectional coupling matrix is ​​incremented by 1 and subjected to a natural logarithm operation, wherein the natural logarithm operation is a logarithmic operation with the constant e as the base and the corresponding element value as the argument, thereby generating a logarithmic coupling matrix; Multiply each element in the logarithmic coupling matrix by the reciprocal of the corresponding target execution segment index value to generate a scaled logarithmic matrix; An exponential function operation is performed on each element in the scaled logarithmic matrix. The exponential function operation is a power operation with a constant e as the base and the corresponding element value as the exponent, to generate an exponential response matrix. Normalize the data in each row of the exponential response matrix by dividing each element in each row by the sum of all elements in the corresponding row to generate a normalized response matrix. Each row of data in the normalized response matrix is ​​multiplied dimension-by-dimensionally with the reciprocal of the corresponding target execution segment index value to generate a position-weighted response matrix; Each row of data in the position-weighted response matrix is ​​accumulated dimension by dimension to generate a penetration response value, which is then arranged according to the target execution segment order to generate a harmonic window penetration sequence.

7. The intelligent dispatch management system for PCB drilling machines based on machine learning according to claim 1, characterized in that, The sequence arrangement module is specifically as follows: Based on the task-equipment adaptation list, the adaptation execution segments corresponding to each PCB drilling task are grouped according to the PCB drilling machine to generate a machine task set. In the set of machine tasks, the adaptation execution segments are grouped according to the aperture level data. Adaptive execution segments with the same aperture level data are grouped into the same group to generate an aperture group execution sequence. In the aperture grouping execution sequence, the adaptation execution segments are grouped according to the aperture distribution area data, and the adaptation execution segments with the same aperture distribution area data are divided into the same group to generate a region grouping execution sequence; In the aforementioned regional grouping execution sequence, the adaptation execution segments are grouped according to the plate thickness segment data. Adaptive execution segments with the same plate thickness segment data are grouped into the same group to generate a plate thickness grouping execution sequence. The adaptation execution segments within each group of the plate thickness grouping execution sequence are sorted from largest to smallest according to their corresponding normalized contribution values ​​to generate a sorted execution sequence within the group. The maximum normalized contribution value within each group is used as the inter-group ranking value, and the groups are sorted from largest to smallest according to the inter-group ranking value to generate an inter-group ranking sequence. Arrange the intra-group sorting execution sequences of each group according to the inter-group sorting sequence to form the task execution order of the corresponding PCB drilling machine. The task execution order of each PCB drilling machine is summarized to form a drilling machine sequence dispatch table.

8. The intelligent dispatch management system for PCB drilling machines based on machine learning according to claim 1, characterized in that, The specific placeholder locking module is: Based on the drilling machine sequence dispatch table, the task execution order corresponding to each PCB drilling machine is extracted, and the corresponding appropriate execution segment is determined one by one according to the task execution order. For each PCB drilling machine, the adaptation execution segment is allocated a continuous time period according to the task execution order to generate a machine occupancy time sequence. Bind each time period in the machine occupancy time sequence to the corresponding adaptive execution segment to generate a machine occupancy mapping relationship; Based on the machine occupancy mapping relationship, a machine occupancy locking operation is performed on each PCB drilling machine. The PCB drilling machine corresponding to the time period within the machine occupancy time sequence is marked as locked; the PCB drilling machine corresponding to the time period not within the machine occupancy time sequence is marked as unlocked. The machine occupancy mapping relationship and machine occupancy lock status are combined to generate a machine occupancy dispatch instruction set.

9. The intelligent dispatch management system for PCB drilling machines based on machine learning according to claim 1, characterized in that, The disturbance optimization module is specifically: Real-time updates of the operating status data of each PCB drilling machine, generating and updating the real-time capability set of the drilling machine; Based on the machine position locking status in the machine position dispatch instruction set, the corresponding adaptation execution section is marked as a reserved section, and the corresponding adaptation execution section is marked as a rearrangement section. According to the task execution order in the drilling machine sequence dispatch table, extract the rearrangement segment corresponding to each PCB drilling machine and arrange them according to the task execution order to generate a local rearrangement segment sequence. For each PCB drilling machine, the penetration-driven Jaya scheduling algorithm is executed. Each rearranged segment in the local rearranged segment sequence is sequentially numbered according to the task execution order. The sequential number is used as the position number, and each rearranged segment is assigned a unique identifier. For each pair of adjacent rearranged segments in the local rearranged segment sequence, an adjacent exchange operation is executed once to generate multiple initial candidate dispatch sequences. The sum of contribution values ​​is obtained by summing the normalized contribution values ​​corresponding to each rearranged segment in the initial candidate dispatch sequence; The normalized value of the spindle vibration amplitude data corresponding to the real-time capability of the updated drilling machine is added to the normalized value of the spindle temperature data to obtain the state penalty value. Subtract the state penalty value from the sum of the contribution values ​​to obtain the sequence evaluation value. The initial candidate assignment sequence with the largest sequence evaluation value is determined as the optimal candidate assignment sequence, and the initial candidate assignment sequence with the smallest sequence evaluation value is determined as the worst candidate assignment sequence. Based on the unique identifier of each rearranged segment, the position number of the corresponding rearranged segment is found in the optimal candidate assignment sequence and the worst candidate assignment sequence, respectively. The difference between the current position number and the corresponding position number in the optimal candidate assignment sequence is calculated and the absolute value is taken to obtain the first position difference. The difference between the current position number and the corresponding position number in the worst candidate assignment sequence is calculated and the absolute value is taken to obtain the second position difference. The current position number is added to the first position difference and the second position difference is subtracted to obtain the updated position number. The rearranged segments are reordered according to the updated position number to generate an updated candidate assignment sequence, and the updated candidate assignment sequence is used as the initial candidate assignment sequence for the next iteration. Repeatedly perform sequence evaluation value calculation, optimal candidate assignment sequence determination, worst candidate assignment sequence determination, and update position number calculation until the maximum number of iterations or the optimal candidate assignment sequence obtained by consecutively setting the number of iterations is consistent. Stop iterating and determine the optimal candidate assignment sequence obtained in the last iteration as the final candidate assignment sequence. The final candidate dispatch sequence is filled into the position corresponding to the machine's non-locked state, while keeping the reserved section position corresponding to the machine's locked state unchanged, thus generating the disturbance-corrected dispatch sequence for the corresponding PCB drilling machine.

10. The intelligent dispatch management system for PCB drilling machines based on machine learning according to claim 1, characterized in that, The handover generation module is specifically: Based on the disturbance correction dispatch sequence of each PCB drilling machine, the task execution order corresponding to each PCB drilling machine is extracted. After each PCB drilling machine completes the adaptation execution section corresponding to the current processing segment, the next adaptation execution section after the current processing segment is selected from the disturbance correction dispatch sequence of the corresponding PCB drilling machine as the next stage task. According to the task execution order of the disturbance correction dispatch sequence, the time connection of the adaptation execution segments of each PCB drilling machine is sequentially performed to form a continuous processing sequence of each PCB drilling machine. The continuous processing sequences of each PCB drilling machine are summarized to generate a continuous processing handover plan.

Images