Double-track cache machine optimization control method and double-track cache machine thereof

Abstract

The invention discloses an optimal control method for a double-track cache machine. The method comprises the following steps: generating a board width heat map according to a board width frequency distribution model; generating a candidate switching window prediction list; forming an optimized input vector; setting plate width constraint, cache constraint and mechanism action constraint; quickly generating a first candidate switching strategy set; calculating a comprehensive performance index of each candidate switching strategy; generating a new candidate switching strategy set to avoid local optimal retention; and the double-track cache machine is driven to complete track mode switching according to the target switching window, and the cache proportion, the lifting layer position, the width adjusting width, the clutch conveying speed and the plate distance are synchronously adjusted according to instructions. According to the method, the track mode switching opportunity can be dynamically optimized according to the incoming material plate width change trend, beat loss, congestion and utilization rate reduction caused by switching lag are effectively avoided, and the flexible scheduling capacity and the production line comprehensive efficiency of the double-track cache machine in the variable production environment are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of double-track, in particular to a double-track buffer machine optimization control method and double-track buffer machine. BACKGROUND

[0002] With the development of electronic manufacturing industry towards multi-variety, small batch and flexibility, the requirements for board width adaptability and production line beat flexibility on surface mount technology production line are increasingly improved. The existing double-track buffer machine is mainly used in the rear section of reflow soldering process to realize product cooling and production beat adjustment. For the complex scene of diversified board width and frequent production demand change, the existing technology generally adopts simple rule-driven and single-threshold switching strategy. When the board width exceeds a certain threshold (such as 300 mm), it is switched to single-track mode production. This mode has the following disadvantages: The board width switching depends on fixed threshold, which fails to dynamically adapt to the change of board width frequency distribution in the production process, resulting in decreased utilization rate of double-track production line, unbalanced load of single-track production line, and frequent beat loss and congestion in the switching window.

[0003] The linkage between multiple actuators lacks global optimization and constraint mechanism, and is often executed in a sequential driving mode, which is prone to action redundancy, increased mechanical impact and rising energy consumption. The existing method generally cannot realize the optimal switching timing and parameter scheduling of the buffer machine in mixed mode, and it is difficult to balance multiple goals such as beat fluctuation suppression, capacity balance and equipment mechanical life, which restricts the intelligent and adaptive control level of the double-track buffer machine in complex production environment. SUMMARY

[0004] One object of the present application is to provide a double-track buffer machine optimization control method. The present application can dynamically optimize the track mode switching timing according to the incoming board width change trend, effectively avoid beat loss, congestion and utilization reduction caused by switching lag, and improve the flexible scheduling capability and overall efficiency of the double-track buffer machine in a variable production environment.

[0005] According to the double-track buffer machine optimization control method of the present application, the method comprises the following steps: S1. Collecting board width statistical data set, performing frequency statistics on the board width statistical data set within a preset sliding time window, constructing a board width frequency distribution model, and generating a board width heat map according to the board width frequency distribution model; S2. Based on the board width frequency distribution model and the board width heat map, a threshold determination rule is used to divide small board width interval and large board width interval, and a candidate switching window prediction list is generated; S3. Obtaining current device state data of the double-track buffer machine, and merging the device state data with the candidate switching window prediction list to form an optimization input vector; S4. Establish a multi-objective optimization function according to the optimization input vector, and set the plate width constraint, the buffer constraint and the mechanism action constraint; S5. Use the trend optimization behavior of the cockroach algorithm to take the currently optimal predicted switching window in the candidate switching window prediction list as the search center, and quickly generate a first candidate switching strategy set; S6. Use the swarm behavior of the cockroach algorithm to search for a second candidate switching strategy set that meets the plate width constraint, the buffer constraint and the mechanism action constraint in the feasible neighborhood of the first candidate switching strategy set, and calculate the comprehensive performance index of each candidate switching strategy through the swarm behavior; S7. Use the dispersion behavior of the cockroach algorithm to trigger random disturbance to adjust the search direction when detecting that the plate width heat map appears mutation or the second candidate switching strategy set falls into local optimum, and generate a new candidate switching strategy set to avoid local optimum retention; S8. Perform comprehensive performance evaluation on the candidate switching strategy set, select the switching strategy with the best comprehensive performance index as the target switching strategy, and issue the target switching strategy to the industrial control system for execution, to drive the double-track buffer machine to complete track mode switching according to the target switching window, and to synchronize adjustment of the buffer occupancy ratio, the lifting layer position, the width adjustment width, the clutch conveying speed and the plate spacing according to the instructions.

[0006] Optionally, the S1 specifically comprises: S11. Collect actual physical width data of the upstream production line at each time point in a continuous time sequence; S12. Extract physical width data collected in a continuous historical time period; S13. Merge the actual physical width data and the historical physical width data to obtain a plate width statistical data set; S14. In the sliding time window length range, for each plate width data in the plate width statistical data set, the number of occurrences of the plate width data in the plate width statistical data set is counted, and the relative occurrence frequency of the plate width data is obtained by the ratio of the number of occurrences to the total number of data in the plate width statistical data set; S15. All plate width data and corresponding relative occurrence frequencies constitute a plate width frequency distribution model; S16. Divide the plate width interval into a plurality of plate width subintervals according to a fixed interval, and count the sum of the relative occurrence frequencies of all plate width data in the plate width statistical data set falling within the subinterval range as the heat value of the plate width subinterval for each plate width subinterval; S17. The heat values of all plate width subintervals and the corresponding plate width intervals jointly draw a plate width heat map.

[0007] Optionally, the S2 specifically comprises: S21. According to the plate width frequency distribution model, interval judgment is performed on each plate width data. If a plate width data is less than or equal to a plate width critical threshold value, the plate width data is classified into a small plate width interval. If a plate width data is greater than the plate width critical threshold value, the plate width data is classified into a large plate width interval; S22. All plate width data in the plate width frequency distribution model is continuously segmented in time sequence. Each time segment is set as a time segment. The type proportion of all plate width data in the time segment is calculated; S23. If the frequency of plate width data belonging to a plate width interval is greater than or equal to a plate width type switching threshold value in any time segment, the corresponding time segment is marked as a candidate switching window; S24. A candidate switching window prediction list is constructed according to all identified time segments; S25. The candidate switching window prediction list is cross-verified with the plate width heat map. If the plate width sub-interval heat value corresponding to a time segment is greater than or equal to a heat critical value, the corresponding time segment is retained as an effective switching window.

[0008] Optionally, the S3 specifically comprises: S31. In-process product position data, track mode state data of the double-track buffer machine at the current time are acquired; S32. Ascending and descending structure posture data at the current time are acquired; S33. Width adjustment mechanism state data at the current time are acquired; S34. The acquired in-process product position data, track mode state data, ascending and descending structure posture data and width adjustment mechanism state data are merged to form a device state data vector; S35. The device state data vector and the final switching window prediction list are structurally fused to construct an optimization input vector.

[0009] Optionally, the S4 specifically comprises: S41. The square of the difference between all production tact values and the production tact average value in the current predicted switching window is summed and divided by the total number of production tact values to obtain a production tact fluctuation target function; S42. The end time of the predicted switching window is subtracted from the start time of the predicted switching window to obtain a switching time target function; S43. The cumulative number of actuator actions in the predicted switching window is counted to obtain an actuator action number target function; S44. The production tact fluctuation target function, the switching time target function and the actuator action number target function are weighted and summed to obtain a multi-objective optimization function; S45. Set a board width constraint, when the track mode state is the double-track dual-purpose mode, the actual physical width of each circuit board is not more than three hundred millimeters, when the track mode state is the single-track large board mode, the actual physical width of each circuit board is not more than four hundred and sixty millimeters; S46. Set a buffer constraint, by judging the work-in-process quantity at all time points in the predicted switching window, only when the work-in-process quantity at each time point is less than or equal to the maximum capacity of the buffer area, the buffer constraint is satisfied; S47. Set a mechanism action constraint, limit the current layer height of all lifting structures between zero and the maximum reachable height of the lifting structure, and limit the current width setting value of all width adjustment mechanisms between the minimum rail spacing and the maximum rail spacing.

[0010] Optionally, the S5 specifically comprises: S51. In the optimization search stage, define the state of each cockroach individual as a three-dimensional parameter vector including the buffer occupancy ratio, the conveying speed and the beat stability coefficient; S52. Set a dynamic guidance function of the optimization search, so that each cockroach individual moves from the current position to the current optimal strategy solution in each iteration according to the combined action of the convergence step length factor and the structure coupling modulation operator; S53. Input the improved state of all cockroach individuals in each iteration to the feasibility checking module, dynamically check the board width constraint, the buffer constraint and the mechanism action constraint, only when the board width is less than or equal to the maximum board width corresponding to the track mode, the work-in-process quantity of the buffer area is less than or equal to the maximum capacity, the current layer height of the lifting structure is between zero and the maximum reachable height, and the current width setting value of the width adjustment mechanism is between the minimum rail spacing and the maximum rail spacing, the cockroach individual is included in the first candidate switching strategy set, otherwise the cockroach individual reverts to the last iteration and updates the direction vector; S54. Repeat the optimization search until the multi-objective optimization function converges or the maximum iteration number is reached, and output the first candidate switching strategy set.

[0011] Optionally, the S6 specifically comprises: S61. Define the first candidate switching strategy set generated by the optimization behavior as a set of strategy solutions; S62. For each first candidate switching strategy solution, perturb in the directions of the buffer occupancy ratio, the conveying speed and the beat stability coefficient respectively, to obtain a set of local perturbation candidate solutions, and all the local perturbation candidate solutions constitute the neighborhood solution set of the first candidate switching strategy solution; S63. Merge all the neighborhood solution sets of the first candidate switching strategy solutions to obtain the second candidate switching strategy set; S64. For each strategy solution in the second candidate switching strategy set, calculate the multi-objective optimization function value of the strategy solution, and further calculate the structural consistency coefficient; S65. For each strategy solution in the second candidate switching strategy set, the multi-objective optimization function value and the structural consistency coefficient are weighted and combined to obtain the cluster comprehensive performance index.

[0012] Optionally, S7 specifically includes: S71. After each optimization iteration cycle, perform differential calculation on two consecutive frames of the board width heat map to obtain the heat change rate; S72. If the rate of change of heat is greater than the threshold for sudden change of heat, or if the improvement of the multi-objective optimization function value of the second candidate switching strategy set in multiple consecutive iterations is less than the convergence threshold, then the second candidate switching strategy set is determined to be trapped in a local optimum. S73. When it is determined that the second candidate switching strategy set is trapped in a local optimum, apply uniformly distributed random perturbations to the buffer ratio, transmission speed and cycle stability coefficient of each strategy solution in the second candidate switching strategy set to obtain a new set of perturbation strategy solutions; S74. Merge all new perturbation strategy solutions to form a new candidate switching strategy set; S75. For each strategy solution in the new candidate switching strategy set, recalculate the multi-objective optimization function value and perform feasibility checks on plate width constraints, buffer constraints, and mechanism action constraints. Only strategy solutions that satisfy all three types of constraints are retained, and the rest are discarded. S76. The retained policy solution set is used as the updated second candidate switching policy set to replace the original policy set that is trapped in local optima and enter the next round of clustering and dispersing behavior iteration.

[0013] A dual-track buffer machine is used to execute a dual-track buffer machine optimization control method, comprising: The central processing unit has a built-in dual-track cache machine optimization control method; The first track buffer frame is cooled by a frame lifting synchronous mechanism controlled by a frame lifting hexagonal transmission rod driven by a frame servo lifting motor. The second track buffer frame is controlled by a two-frame servo lifting motor that drives the two-frame lifting hexagonal transmission rod to control the two-frame lifting synchronous mechanism for cooling the second product. The first and second track cache frames together form a dual-track cache frame.

[0014] The beneficial effects of this invention are: This invention constructs a real-time board width frequency distribution model and heat map, and introduces a sliding window statistics and threshold judgment mechanism into the cockroach algorithm to intelligently analyze the dynamic distribution of different board widths during the production process. This enables proactive prediction and fine control of the switching window, and can dynamically optimize the timing of track mode switching based on the changing trend of incoming board width. This effectively avoids cycle time loss, congestion, and reduced utilization caused by switching lag, and improves the flexible scheduling capability and overall production line efficiency of the dual-track buffer machine in a variable production environment.

[0015] This invention introduces three-dimensional parameters—buffer ratio, delivery speed, and cycle stability coefficient—into the optimization and swarming behavior of the cockroach algorithm. It establishes a structural coupling modulation operator and a structural consistency coefficient to dynamically adjust the search direction and stride of individuals during the optimization iteration process. This enables the cockroach algorithm to adaptively avoid physical bottlenecks such as full buffer, speed fluctuations, and cycle anomalies while approaching the global optimum. It effectively compresses redundant action sequences, reduces the repetition and mechanical impact of lifting and width adjustment actions, and improves the mechanical friendliness and energy efficiency of equipment operation.

[0016] In the distributed behavior stage, the invention utilizes heatmap mutation detection and strategy convergence monitoring. When the board width distribution or strategy set is detected to be trapped in a local optimum, multi-dimensional perturbations are automatically applied to the buffer ratio, conveying speed, and cycle stability coefficient, dynamically generating a new candidate switching strategy set. This effectively avoids the local stickiness and strategy failure phenomena that occur in traditional algorithms under complex working conditions, ensuring stable tracking of buffer release and production line cycle under production environment fluctuations and abnormal disturbances. It achieves highly reliable online optimization of multiple mechanisms, multiple parameters, and multiple objectives. Attached Figure Description

[0017] 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 invention presents an optimized control method for a dual-track buffer machine and a flowchart of the dual-track buffer machine. Detailed Implementation

[0018] Example 1: Reference Figure 1 A dual-track buffer machine optimization control method, comprising: S1. Collect a set of board width statistics data, perform frequency statistics on the set of board width statistics data within a preset sliding time window, construct a board width frequency distribution model, and generate a board width heat map based on the board width frequency distribution model; In this embodiment, S1 specifically includes: S11. Collect the actual physical width data of the upstream production line at each time point in a continuous time series; All actual physical width data form a continuous actual physical width data sequence. The actual physical width data is used to describe the physical width of a single circuit board at each acquisition time. The board width range is A: 50~300; range B: 50~460.

[0019] S12. Extract physical width data collected within a continuous historical time period; The physical width data for all historical time periods constitutes the historical physical width data sequence. The historical physical width data is used to supplement the actual physical width data sequence to reflect all possible board width distributions within the target time period.

[0020] S13. Merge the actual physical width data with the historical physical width data to obtain the board width statistical data set; The board width statistics set is a collection of all actual physical width data and historical physical width data within the target time period. The board width statistics set describes all possible board width distribution scenarios within the target time period.

[0021] S14. Within the length of the sliding time window, for each piece of board width data in the board width statistical data set, count the number of times the board width data appears in the board width statistical data set, and obtain the relative frequency of occurrence of the board width data by the ratio of the number of occurrences to the total number of data in the board width statistical data set; The relative frequency of occurrence of all plate width data collectively characterizes the probability distribution of each plate width data within the statistical interval.

[0022] S15. All board width data and their corresponding relative occurrence frequencies constitute a board width frequency distribution model; Each plate width data point corresponds one-to-one with its relative frequency of occurrence. The plate width frequency distribution model is used to represent the popularity level and distribution characteristics of all plate widths within the target time period.

[0023] S16. Divide the plate width interval into several plate width sub-intervals according to a fixed interval. For each plate width sub-interval, calculate the sum of the relative occurrence frequencies of all plate width data falling within the range of the sub-interval in the plate width statistical data set, and use it as the heat value of the plate width sub-interval. The heat value is used to measure the occurrence density of each plate width sub-interval within the target time period.

[0024] S17. The heat values ​​of all plate width sub-intervals are used to plot the plate width heat map together with the corresponding plate width intervals.

[0025] The heat values ​​of all board width sub-intervals are mapped one-to-one with the corresponding board width sub-interval ranges. The board width heat map is plotted with board width on the horizontal axis and heat values ​​on the vertical axis to describe the frequency change trend of circuit boards with different width ranges within the target time period.

[0026] S2. Based on the board width frequency distribution model and board width heat map, a threshold judgment rule is used to divide the small board width interval and the large board width interval, and a candidate switching window prediction list is generated; In this embodiment, S2 specifically includes: S21. Based on the board width frequency distribution model, each board width data is judged in intervals. If a certain board width data is less than or equal to the board width critical threshold, the board width data is classified as a small board width interval. If a certain board width data is greater than the board width critical threshold, the board width data is classified as a large board width interval. In the process of determining the interval for each board width data, it is necessary to set a critical threshold for board width. The critical threshold for board width is used to divide all board width data into small board width intervals and large board width intervals. In Example 1, the critical threshold for board width is 300mm.

[0027] S22. Divide all board width data in the board width frequency distribution model into continuous segments according to time order, set each time segment as a time period, and calculate the proportion of the type of all board width data in that time period; Each time period includes a start time and an end time, and belongs to the type of small board width interval or large board width interval.

[0028] S23. If, within any time period, the frequency of board width data belonging to a certain board width range is greater than or equal to the board width type switching threshold, then the corresponding time period is marked as a candidate switching window. The board width type switching threshold is used to determine the dominant trend of a certain board width range within a time period. The proportion of the type is obtained by the ratio of the number of board width data of a certain type within the corresponding time period to the total number of all board width data within the corresponding time period.

[0029] S24. Construct a candidate switching window prediction list based on all identified time periods; Each element in the candidate switching window prediction list includes the start time of the time period, the end time of the time period, and the dominant board width interval type of the time period, which is either a small board width interval or a large board width interval.

[0030] S25. Cross-validate the candidate switching window prediction list with the board width heat map. If the heat value of the board width sub-interval corresponding to a certain time period is greater than or equal to the heat critical value, then retain the corresponding time period as a valid switching window.

[0031] The heat threshold is used to eliminate inactive intervals with low heat fluctuations, ensuring the dynamic effectiveness of candidate switching windows.

[0032] S3. Obtain the current device status data of the dual-track buffer machine, and merge the device status data with the candidate switching window prediction list to form an optimized input vector; In this embodiment, S3 specifically includes: S31. Obtain the current work-in-process position data and track mode status data of the dual-track buffer machine; Each work-in-process item in the work-in-process number set corresponds to a spatial coordinate. The work-in-process position data describes the linear position of all work-in-process items in the track direction and the level number in the cache layer direction. The set of spatial coordinates of all work-in-process items represents the spatial distribution status of all work-in-process items in the current cache area.

[0033] The track mode status data indicates whether the equipment track structure is currently in dual-track dual-use mode or single-track large plate mode. Dual-track dual-use mode means that the maximum supported plate width does not exceed 300 mm, while single-track large plate mode means that the maximum supported plate width does not exceed 460 mm.

[0034] S32. Obtain the current attitude data of the lifting structure; Each lifting structure in the set of lifting structure numbers corresponds to a current level height. The lifting structure attitude data is used to describe the current height distribution of all lifting structures. The level height of each lifting structure is used to reflect the current buffer status and reachability status of the current buffer track in the vertical direction.

[0035] S33. Obtain the current status data of the width adjustment mechanism; Each width adjustment mechanism in the set of adjustment mechanism numbers corresponds to a current width setting value. The width adjustment mechanism status data is used to describe the current guide rail spacing distribution of all width adjustment mechanisms. The current width setting value of each width adjustment mechanism is used to reflect the current minimum passable plate width limit.

[0036] S34. Merge the acquired work-in-process position data, track mode status data, lifting structure attitude data, and width adjustment mechanism status data to form an equipment status data vector; The device status data vector is used to comprehensively describe all the operational status characteristics of the current dual-track buffer machine.

[0037] S35. Structurally fuse the device status data vector with the final switching window prediction list to construct an optimized input vector.

[0038] In Example 1, the work-in-process position data, track mode status data, lifting structure attitude data, and width adjustment mechanism status data in the equipment status data vector are numbered and associated with timestamps, and are corresponding one-to-one with the start time, end time, and main plate width interval type of each switching window in the final switching window prediction list. The equipment status data within the same time period are combined with the switching windows of the time period, and the combination results within all time periods are arranged sequentially to obtain a fusion dataset arranged in chronological order. Each set of equipment status data in the fusion dataset and its corresponding switching window information are used as a whole input, and all whole inputs are combined sequentially to form the optimized input vector.

[0039] S4. Establish a multi-objective optimization function based on the optimized input vector, and set plate width constraints, buffer constraints, and mechanism motion constraints; In this embodiment, S4 specifically includes: S41. The objective function for production cycle fluctuation is obtained by summing the squares of the differences between all production cycle values ​​and the average production cycle value within the current prediction switching window and dividing the sum by the total number of production cycle values. S42. The objective function for switching time is obtained by subtracting the predicted start time of the switching window from the predicted end time of the switching window; The switching time objective function is used to represent the total time required to complete one track mode switch.

[0040] S43. Obtain the objective function for the number of actuator actions by counting the cumulative number of actuator actions within the prediction switching window; The actuator action represents the total number of all lifting, width adjustment, and clutch switching actions within the prediction switching window.

[0041] S44. By weighting and summing the objective functions of production cycle fluctuation, switching time, and the number of actuator actions, a multi-objective optimization function is obtained; The multi-objective optimization function is used to comprehensively minimize three objectives: production cycle fluctuation, switching time, and the number of actuator actions.

[0042] S45. Set board width constraints: when the track mode is dual-track dual-use mode, the actual physical width of each board shall not exceed 300 mm; when the track mode is single-track large board mode, the actual physical width of each board shall not exceed 460 mm. Board width constraints are used to ensure that the actual physical width of each board does not exceed the maximum allowable width value in the current track mode.

[0043] S46. Set cache constraints. By judging the number of work-in-process at all time points within the prediction switching window, the cache constraints are only satisfied when the number of work-in-process at each time point is less than or equal to the maximum capacity of the cache. Cache constraints are used to ensure that the number of works in the cache at any given time does not exceed the maximum capacity of the cache.

[0044] S47. Set mechanism motion constraints to limit the current floor height of all lifting structures to between zero and the maximum reachable height of the lifting structure, and the current width setting value of all width adjustment mechanisms to between the minimum guide rail spacing and the maximum guide rail spacing.

[0045] The objective functions of production cycle fluctuation, switching time, and number of actuator actions, along with plate width constraints, buffer constraints, and mechanism action constraints, are used as evaluation criteria to form a complete multi-objective optimization function template.

[0046] S5. Utilizing the optimization behavior of the cockroach algorithm, the first candidate switching strategy set is quickly generated by taking the currently optimal predicted switching window in the candidate switching window prediction list as the search center. In this embodiment, S5 specifically includes: S51. In the optimization search phase, the state of each individual cockroach is defined as a three-dimensional parameter vector including buffer ratio, delivery speed and cycle stability coefficient; The buffer ratio is used to represent the ratio of the real-time loading capacity of the buffer area to the maximum capacity. The conveying speed is used to reflect the movement rate of the clutch conveyor mechanism. The cycle stability coefficient is used to describe the adaptive ability of the track buffer machine to reflow welding cycle fluctuations.

[0047] S52. Set a dynamic guiding function for the optimization search so that each individual cockroach moves from its current position toward the current optimal strategy solution in each iteration according to the combined effect of the convergence step size factor and the structural coupling modulation operator. The convergence step size factor is used to approach stability control as the number of iterations decreases. The structural coupling modulation operator is a nonlinear function describing the constraints of the dual-track buffer mechanism. The structural coupling modulation operator is used to perform weighted normalization and superposition correction on the differences between the current cockroach individual's buffer ratio and the current window buffer utilization rate, the current conveying speed and the current window conveying speed, and the current window cycle stability coefficient. When the buffer area approaches full load, the current window conveying speed deviates from the target value, or the cycle fluctuation intensifies, the value of the structural coupling modulation operator decreases, limiting the stride of the cockroach individual to quickly approach the optimal solution, thereby preventing overshoot or mechanical shock caused by equipment inertia.

[0048]

[0049] in, Let represent the current state vector of the k-th cockroach in the cockroach algorithm. This represents the state vector of the k-th cockroach in the cockroach algorithm after one iteration. The multi-objective optimization function at the t-th iteration The smallest individual cockroach state, α(t) is the time decay factor, with a value range of 0 < α(t) ≤ 1, used to approach stability control as the number of iterations decreases. For structural coupling modulation operators, describe the nonlinear function of the constraints of the dual-track buffer mechanism: ; in, For modulation weighting factors, and These represent the maximum utilization of the buffer and the maximum transfer speed, respectively. The buffer percentage modulation weighting factor is used to adjust the influence of the buffer percentage term on the modulation operator; it is dimensionless. individual cockroaches The cache percentage corresponding to the current strategy is expressed as the ratio of the current cache load to the maximum capacity. Let be the cache percentage corresponding to the optimal solution in the t-th iteration. This serves as the normalized baseline value for cache percentage. individual cockroaches The delivery speed corresponding to the current strategy, Let the optimal solution in the t-th iteration be the transport speed. The maximum allowable conveying speed of the system. individual cockroaches The cycle stability coefficient corresponding to the current strategy reflects the adaptive capability of the track buffer machine to production line cycle fluctuations. Let be the cycle stability coefficient corresponding to the optimal solution in the t-th iteration.

[0050] The dynamic guiding function of the optimization search is used to describe the degree of optimization coupling of individual cockroaches under the current buffer utilization, conveying speed and cycle stability conditions. When the buffer is close to full load, the conveying speed deviates from the target value or the cycle fluctuation intensifies, the structural coupling modulation operator automatically decreases, limiting the stride of individual cockroaches to the optimal solution, thereby preventing over-adjustment or mechanical shock caused by equipment inertia.

[0051] S53. Input the improved state of all cockroach individuals in each iteration into the feasibility verification module to dynamically verify the plate width constraint, buffer constraint and mechanism action constraint. Only when the plate width is less than or equal to the maximum plate width corresponding to the track mode, the number of work-in-process in the buffer area is less than or equal to the maximum capacity, the current layer height of the lifting structure is between zero and the maximum reachable height, and the current width setting value of the width adjustment mechanism is between the minimum guide rail spacing and the maximum guide rail spacing, is the cockroach individual included in the first candidate switching strategy set. Otherwise, the cockroach individual will regress to the previous iteration and update the direction vector. S54. Repeat the optimization search until the multi-objective optimization function converges or the maximum number of iterations is reached, then terminate the iteration and output the first candidate switching strategy set.

[0052] Each strategy solution satisfies the requirement of minimizing the multi-objective optimization function and is consistent with the characteristics of buffer state, delivery speed and cycle stability.

[0053] S6. Utilize the clustering behavior of the cockroach algorithm to search for a second candidate switching strategy set that satisfies the plate width constraint, buffer constraint, and mechanism action constraint within the feasible neighborhood of the first candidate switching strategy set, and calculate the comprehensive performance index of each candidate switching strategy through the clustering behavior. In this embodiment, S6 specifically includes: S61. Define the first candidate switching strategy set generated by the optimism behavior as a set of strategy solutions; S62. For each first candidate switching strategy solution, perturb the buffer ratio, transmission speed and cycle stability coefficient respectively to obtain a set of local perturbation candidate solutions. All local perturbation candidate solutions constitute the neighborhood solution set of the first candidate switching strategy solution. S63. Merge the neighborhood solution sets of all first candidate switching strategy solutions to obtain the second candidate switching strategy set; S64. For each strategy solution in the second candidate switching strategy set, calculate the multi-objective optimization function value of the strategy solution, and further calculate the structural consistency coefficient; In Example 1, the structural consistency coefficient is calculated as follows: the difference between the cache ratio of the strategy solution and the cache ratio in the current window optimal state is divided by the maximum utilization rate of the cache area; the difference between the transmission speed of the strategy solution and the transmission speed in the current window optimal state is divided by the maximum allowable transmission speed of the system; and the absolute value of the difference between the clock stability coefficient of the strategy solution and the clock stability coefficient in the current window optimal state is taken directly, and then weighted summation is performed. The structural consistency coefficient is used to quantify the synergy between the strategy solution and the current physical state.

[0054] S65. For each strategy solution in the second candidate switching strategy set, the multi-objective optimization function value and the structural consistency coefficient are weighted and combined to obtain the cluster comprehensive performance index.

[0055] The cluster comprehensive performance index is used to simultaneously reflect the multi-objective optimization effect and structural consistency of the strategy solution. The smaller the value of the cluster comprehensive performance index, the better the comprehensive performance of the strategy solution.

[0056] S7. Utilizing the dispersed behavior of the cockroach algorithm, when a sudden change is detected in the board width heatmap or the second candidate switching strategy set gets stuck in a local optimum, a random perturbation is triggered to adjust the search direction and generate a new candidate switching strategy set to avoid getting stuck in a local optimum. In this embodiment, S7 specifically includes: S71. After each optimization iteration cycle, perform differential calculation on two consecutive frames of the board width heat map to obtain the heat change rate; The heat change rate is obtained by taking the absolute value of the difference between the heat value of the current frame board width sub-interval and the heat value of the previous frame board width sub-interval and averaging them. The heat change rate is used to measure the degree of abrupt change in the board width frequency distribution.

[0057] S72. If the rate of change of heat is greater than the threshold for sudden change of heat, or if the improvement of the multi-objective optimization function value of the second candidate switching strategy set in multiple consecutive iterations is less than the convergence threshold, then the second candidate switching strategy set is determined to be trapped in a local optimum. S73. When it is determined that the second candidate switching strategy set is trapped in a local optimum, apply uniformly distributed random perturbations to the buffer ratio, transmission speed and cycle stability coefficient of each strategy solution in the second candidate switching strategy set to obtain a new set of perturbation strategy solutions; The disturbance amplitude is set according to a fixed proportion of the allowable range of each parameter.

[0058] S74. Merge all new perturbation strategy solutions to form a new candidate switching strategy set; S75. For each strategy solution in the new candidate switching strategy set, recalculate the multi-objective optimization function value and perform feasibility checks on plate width constraints, buffer constraints, and mechanism action constraints. Only strategy solutions that satisfy all three types of constraints are retained, and the rest are discarded. S76. The retained policy solution set is used as the updated second candidate switching policy set to replace the original policy set that is trapped in local optima and enter the next round of clustering and dispersing behavior iteration.

[0059] S8. Perform a comprehensive performance evaluation on the candidate switching strategy set, select the switching strategy with the best comprehensive performance index as the target switching strategy, send the target switching strategy to the industrial control system for execution, drive the dual-track buffer machine to complete the track mode switching according to the target switching window, and synchronously adjust the buffer ratio, lifting and lowering layer, width adjustment, clutch conveying speed and board spacing according to the instructions.

[0060] A dual-track buffer machine is used to execute a dual-track buffer machine optimization control method, comprising: The central processing unit has a built-in dual-track cache machine optimization control method; The first track buffer frame is cooled by a frame lifting synchronous mechanism controlled by a frame lifting hexagonal transmission rod driven by a frame servo lifting motor. The second track buffer frame is controlled by a two-frame servo lifting motor that drives the two-frame lifting hexagonal transmission rod to control the two-frame lifting synchronous mechanism for cooling the second product. The first and second track cache frames together form a dual-track cache frame.

[0061] Example 2: In a certain production batch, the width sequence (unit: mm) of the following 30 circuit boards was collected in real time at the upstream SMT reflow soldering outlet of the dual-track buffer machine: 62, 67, 68, 75, 92, 110, 245, 262, 297, 299, 305, 316, 442, 275, 278, 294, 76, 78, 92, 115, 283, 290, 443, 455, 460, 110, 75, 67, 67, 285. The data synchronization interval was 4 seconds per board, and the actual production time was approximately 2 minutes.

[0062] The system first integrates historical data from the past two hours with the aforementioned real-time data based on a sliding window (window length 120 seconds) to form a statistical set of board width data. The statistical results show that the frequency of widths in the 60-120 mm range is 38%, the 250-300 mm range is 28%, the 400-460 mm range is 16%, and the remaining ranges account for 18% in total. The real-time heat map shows that the frequency of large board widths increases significantly from the 10th board (width 299 mm) to the 13th board (width 442 mm).

[0063] The algorithm determines that boards with a width ≤300mm are in the small board width range, and those with a width greater than 300mm are in the large board width range. Combined with real-time heat map changes, boards 10-13 show a "dominant range switching window". When the 9th board is cached, the cache utilization rate reaches 92% (27 / 30). The track mode is dual-track dual-use, the lifting structure layer is 350mm and 200mm, and the current width adjustment mechanism is 70mm and 75mm respectively.

[0064] The system uses the period around the arrival of the 10th board as the prediction switching window, with a start time of 39 seconds, an end time of 57 seconds, and a window length of 18 seconds. The dominant interval type is large boards. Within this window, the upstream surface mount process feedback cycle time is 4 seconds per board, the downstream testing equipment feedback cycle time is stable, the actual average output cycle time of the buffer machine is 4.02 seconds, and the standard deviation of the fluctuation is 0.11 seconds.

[0065] The algorithm automatically integrates the above data to construct an optimized input vector and establish a multi-objective optimization function. The previous round of switching actions consisted of 5 actions (2 width adjustments, 1 elevation change, and 2 clutch engagements). The initial objective for this round is 1 width adjustment, 1 elevation change, and 1 clutch engagement, with the beat fluctuation target controlled within 0.10 seconds.

[0066] In the optimization phase, the system initializes 10 groups of cockroaches, each with a state of (buffer percentage, conveying speed, and cycle stability coefficient). The initial values ​​for each group are (0.92, 520 mm / s, 0.96), (0.91, 515 mm / s, 0.95), ... up to (0.86, 505 mm / s, 0.94). The convergence step size is 0.7, and the dynamic variation range of the structural coupling modulation operator is 0.68~1. After the first round, all individuals approach the optimal solution (0.89, 525 mm / s, 0.98), the buffer percentage decreases, the exit speed slightly increases, and the cycle stability improves.

[0067] During the clustering phase, for the optimal solution, perturbations of ±2%, ±10mm / s, and ±0.01 were applied to three parameters: buffer ratio, conveying speed, and cycle time stability. Thirty neighborhood solutions were obtained, all of which underwent constraint verification. Verification results: 26 solutions satisfied physical constraints such as board width, buffer, and mechanism movement; 4 solutions were eliminated because the buffer ratio exceeded 95%. A structural consistency coefficient was calculated for each solution. In Example 2, it was (0.88, 530mm / s, 0.99), with a total difference of 0.017 from the current window's optimal parameters, resulting in a weighted consistency coefficient of 0.96. The solution with the lowest overall performance index was (0.89, 525mm / s, 0.98), corresponding to an action sequence of 1 width adjustment, 1 lifting / lowering, and 1 clutch engagement / disengagement. The estimated total switching action time was 11 seconds, with a production line exit cycle time fluctuation of 0.08 seconds, reducing the number of actions by 2 and the fluctuation by 0.05 seconds compared to previous solutions.

[0068] Upon entering the distributed behavior phase, the widths of boards 13-16 were detected to be 442mm, 443mm, 455mm, and 460mm, respectively, with a heatmap change rate of 0.19, exceeding the heatmap abrupt change threshold of 0.15. The system determined that the switching strategy set might be trapped in a local optimum. Immediately, ±3%, ±15mm / s, and ±0.03 perturbations were applied to the buffer ratio, delivery speed, and cycle time stability of the current 26 strategy solutions, generating a new candidate strategy set. After constraint verification, 23 solutions passed. Further evaluation of the comprehensive performance indicators revealed that the optimal solution was adjusted to (0.86, 538mm / s, 0.97). The action sequence remained the same, but the switching window was advanced to 0.9 seconds before the 12th board, reducing cycle time jitter to 0.07 seconds.

[0069] Full-process simulation data shows that during batch switching in mixed production lines, the average buffer utilization rate increased to 94%, the average number of switching actions decreased from 4.9 to 2.7, and the mechanical impact measurement (peak acceleration) decreased from 1.33g to 0.81g. Within all switching windows, the maximum fluctuation in the exit cycle time was less than 0.12 seconds, and no buffer congestion or blockage occurred. For occasional downtime in downstream placement processes, the system automatically postponed the optimal switching window to 4.5 seconds before the arrival of the next large board, adapting to cycle time changes in real time. No manual intervention or abnormal line stoppages occurred during three consecutive hours of production.

[0070] 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 dual-track buffer machine optimization control method, characterized in that, include: S1. Collect a set of board width statistics data, perform frequency statistics on the set of board width statistics data within a preset sliding time window, construct a board width frequency distribution model, and generate a board width heat map based on the board width frequency distribution model; S2. Based on the board width frequency distribution model and board width heat map, a threshold judgment rule is used to divide the small board width interval and the large board width interval, and a candidate switching window prediction list is generated; S3. Obtain the current device status data of the dual-track buffer machine, and merge the device status data with the candidate switching window prediction list to form an optimized input vector; S4. Establish a multi-objective optimization function based on the optimized input vector, and set plate width constraints, buffer constraints, and mechanism motion constraints; S5. Utilizing the optimization behavior of the cockroach algorithm, the first candidate switching strategy set is quickly generated by taking the currently optimal predicted switching window in the candidate switching window prediction list as the search center. S6. Utilize the clustering behavior of the cockroach algorithm to search for a second candidate switching strategy set that satisfies the plate width constraint, buffer constraint, and mechanism action constraint within the feasible neighborhood of the first candidate switching strategy set, and calculate the comprehensive performance index of each candidate switching strategy through the clustering behavior. S7. Utilizing the dispersed behavior of the cockroach algorithm, when a sudden change is detected in the board width heatmap or the second candidate switching strategy set gets stuck in a local optimum, a random perturbation is triggered to adjust the search direction and generate a new candidate switching strategy set to avoid getting stuck in a local optimum. S8. Perform a comprehensive performance evaluation on the candidate switching strategy set, select the switching strategy with the best comprehensive performance index as the target switching strategy, send the target switching strategy to the industrial control system for execution, drive the dual-track buffer machine to complete the track mode switching according to the target switching window, and synchronously adjust the buffer ratio, lifting and lowering layer, width adjustment, clutch conveying speed and board spacing according to the instructions.

2. The dual-track buffer machine optimization control method according to claim 1, characterized in that, S1 specifically includes: S11. Collect the actual physical width data of the upstream production line at each time point in a continuous time series; S12. Extract physical width data collected within a continuous historical time period; S13. Merge the actual physical width data with the historical physical width data to obtain the board width statistical data set; S14. Within the length of the sliding time window, for each piece of board width data in the board width statistical data set, count the number of times the board width data appears in the board width statistical data set, and obtain the relative frequency of occurrence of the board width data by the ratio of the number of occurrences to the total number of data in the board width statistical data set; S15. All board width data and their corresponding relative occurrence frequencies constitute a board width frequency distribution model; S16. Divide the plate width interval into several plate width sub-intervals according to a fixed interval. For each plate width sub-interval, calculate the sum of the relative occurrence frequencies of all plate width data falling within the range of the sub-interval in the plate width statistical data set, and use it as the heat value of the plate width sub-interval. S17. The heat values ​​of all plate width sub-intervals are used to plot the plate width heat map together with the corresponding plate width intervals.

3. The dual-track buffer machine optimization control method according to claim 1, characterized in that, S2 specifically includes: S21. Based on the board width frequency distribution model, each board width data is judged in intervals. If a certain board width data is less than or equal to the board width critical threshold, the board width data is classified as a small board width interval. If a certain board width data is greater than the board width critical threshold, the board width data is classified as a large board width interval. S22. Divide all board width data in the board width frequency distribution model into continuous segments according to time order, set each time segment as a time period, and calculate the proportion of the type of all board width data in that time period; S23. If, within any time period, the frequency of board width data belonging to a certain board width range is greater than or equal to the board width type switching threshold, then the corresponding time period is marked as a candidate switching window. S24. Construct a candidate switching window prediction list based on all identified time periods; S25. Cross-validate the candidate switching window prediction list with the board width heat map. If the heat value of the board width sub-interval corresponding to a certain time period is greater than or equal to the heat critical value, then retain the corresponding time period as a valid switching window.

4. The dual-track buffer machine optimization control method according to claim 1, characterized in that, S3 specifically includes: S31. Obtain the current work-in-process position data and track mode status data of the dual-track buffer machine; S32. Obtain the current attitude data of the lifting structure; S33. Obtain the current status data of the width adjustment mechanism; S34. Merge the acquired work-in-process position data, track mode status data, lifting structure attitude data, and width adjustment mechanism status data to form an equipment status data vector; S35. Structurally fuse the device status data vector with the final switching window prediction list to construct an optimized input vector.

5. The dual-track buffer machine optimization control method according to claim 1, characterized in that, S4 specifically includes: S41. The objective function for production cycle fluctuation is obtained by summing the squares of the differences between all production cycle values ​​and the average production cycle value within the current prediction switching window and dividing the sum by the total number of production cycle values. S42. The objective function for switching time is obtained by subtracting the predicted start time of the switching window from the predicted end time of the switching window; S43. Obtain the objective function for the number of actuator actions by counting the cumulative number of actuator actions within the prediction switching window; S44. By weighting and summing the objective functions of production cycle fluctuation, switching time, and the number of actuator actions, a multi-objective optimization function is obtained; S45. Set board width constraints: when the track mode is dual-track dual-use mode, the actual physical width of each board shall not exceed 300 mm; when the track mode is single-track large board mode, the actual physical width of each board shall not exceed 460 mm. S46. Set cache constraints. By judging the number of work-in-process at all time points within the prediction switching window, the cache constraints are only satisfied when the number of work-in-process at each time point is less than or equal to the maximum capacity of the cache. S47. Set mechanism motion constraints to limit the current floor height of all lifting structures to between zero and the maximum reachable height of the lifting structure, and the current width setting value of all width adjustment mechanisms to between the minimum guide rail spacing and the maximum guide rail spacing.

6. The dual-track buffer machine optimization control method according to claim 1, characterized in that, S5 specifically includes: S51. In the optimization search phase, the state of each individual cockroach is defined as a three-dimensional parameter vector including buffer ratio, delivery speed and cycle stability coefficient; S52. Set a dynamic guiding function for the optimization search so that each individual cockroach moves from its current position toward the current optimal strategy solution in each iteration according to the combined effect of the convergence step size factor and the structural coupling modulation operator. S53. Input the improved state of all cockroach individuals in each iteration into the feasibility verification module to dynamically verify the plate width constraint, buffer constraint and mechanism action constraint. Only when the plate width is less than or equal to the maximum plate width corresponding to the track mode, the number of work-in-process in the buffer area is less than or equal to the maximum capacity, the current layer height of the lifting structure is between zero and the maximum reachable height, and the current width setting value of the width adjustment mechanism is between the minimum guide rail spacing and the maximum guide rail spacing, is the cockroach individual included in the first candidate switching strategy set. Otherwise, the cockroach individual will regress to the previous iteration and update the direction vector. S54. Repeat the optimization search until the multi-objective optimization function converges or the maximum number of iterations is reached, then terminate the iteration and output the first candidate switching strategy set.

7. The dual-track buffer machine optimization control method according to claim 1, characterized in that, S6 specifically includes: S61. Define the first candidate switching strategy set generated by the optimism behavior as a set of strategy solutions; S62. For each first candidate switching strategy solution, perturb the buffer ratio, transmission speed and cycle stability coefficient respectively to obtain a set of local perturbation candidate solutions. All local perturbation candidate solutions constitute the neighborhood solution set of the first candidate switching strategy solution. S63. Merge the neighborhood solution sets of all first candidate switching strategy solutions to obtain the second candidate switching strategy set; S64. For each strategy solution in the second candidate switching strategy set, calculate the multi-objective optimization function value of the strategy solution, and further calculate the structural consistency coefficient; S65. For each strategy solution in the second candidate switching strategy set, the multi-objective optimization function value and the structural consistency coefficient are weighted and combined to obtain the cluster comprehensive performance index.

8. The dual-track buffer machine optimization control method according to claim 1, characterized in that, Specifically, S7 includes: S71. After each optimization iteration cycle, perform differential calculation on two consecutive frames of the board width heat map to obtain the heat change rate; S72. If the rate of change of heat is greater than the threshold for sudden change of heat, or if the improvement of the multi-objective optimization function value of the second candidate switching strategy set in multiple consecutive iterations is less than the convergence threshold, then the second candidate switching strategy set is determined to be trapped in a local optimum. S73. When it is determined that the second candidate switching strategy set is trapped in a local optimum, apply uniformly distributed random perturbations to the buffer ratio, transmission speed and cycle stability coefficient of each strategy solution in the second candidate switching strategy set to obtain a new set of perturbation strategy solutions; S74. Merge all new perturbation strategy solutions to form a new candidate switching strategy set; S75. For each strategy solution in the new candidate switching strategy set, recalculate the multi-objective optimization function value and perform feasibility checks on plate width constraints, buffer constraints, and mechanism action constraints. Only strategy solutions that satisfy all three types of constraints are retained, and the rest are discarded. S76. The retained policy solution set is used as the updated second candidate switching policy set to replace the original policy set that is trapped in local optima and enter the next round of clustering and dispersing behavior iteration.

9. A dual-track buffer machine, used to execute the dual-track buffer machine optimization control method according to any one of claims 1-8, characterized in that, include: The central processing unit has a built-in dual-track cache machine optimization control method; The first track buffer frame is cooled by a frame lifting synchronous mechanism controlled by a frame lifting hexagonal transmission rod driven by a frame servo lifting motor. The second track buffer frame is controlled by a two-frame servo lifting motor that drives the two-frame lifting hexagonal transmission rod to control the two-frame lifting synchronous mechanism for cooling the second product. The first and second track cache frames together form a dual-track cache frame.

Images