PCB production real-time scheduling method and system based on hybrid intelligent decision

Abstract

The invention provides a PCB production real-time scheduling method and system based on hybrid intelligent decision, and the method comprises the steps: obtaining the production information of PCB production, and constructing a corresponding digital twinborn for the production information; identifying dominant process tasks and implicit process tasks in the production work order based on the state of digital twinning; obtaining a current service strategy and loading a hard constraint rule set and a soft constraint rule set based on the current service strategy; integrating the dominant process task, the implicit process task, the hard constraint rule set and the soft constraint rule set to obtain a scheduling problem instance; converting the scheduling problem instance into a target function, and performing multi-target optimization solution on the target function to output an initial scheduling plan; and real-time monitoring and dynamic increment rescheduling circulation are carried out on the initial scheduling plan to complete PCB production real-time scheduling, the field expert rule, the operational research optimization algorithm and real-time data feedback are deeply fused in a closed-loop system, and dynamic self-adaptive optimization of the production scheduling is realized.

Description

Technical Field

[0001] This invention belongs to the technical field of industrial automation and information technology, specifically relating to a real-time scheduling method and system for PCB production based on hybrid intelligent decision-making. Background Technology

[0002] In the manufacturing sector, production scheduling is a core element. Its purpose is to allocate production tasks (work orders) to production resources (machines) while satisfying a series of constraints (such as delivery time, resources, and processes), in order to achieve one or more optimization goals (such as shortest production cycle, highest equipment utilization, and lowest production cost). The technical means to achieve scheduling typically include: Manufacturing Execution System (MES): Primarily responsible for the execution, monitoring, and management of the production process. Some MES systems include simple scheduling modules, but their core focus is on "execution" rather than "planning." Their scheduling functions are usually based on local, pre-defined dispatch rules (such as FIFO, priority for urgent work orders), lacking a holistic view and predictive ability of the overall factory status. Therefore, their decisions are often local and suboptimal.

[0003] Enterprise Resource Planning (ERP): Focuses on macro-level resource planning at the enterprise level, such as order management and Material Requirements Planning (MRP). Its scheduling function is usually a coarse-grained plan with "unlimited capacity" or a bottleneck process assessment based on "coarse-grained capacity planning (RCCP)". It cannot handle the real-time constraints and dynamic changes of hundreds of specific resources such as machines, tooling, and personnel in the shop floor.

[0004] Advanced Planning and Scheduling (APS) Systems: Professional planning and scheduling software that bridges the gap between ERP macro-planning and MES shop floor execution. These systems typically use operations research algorithms such as mathematical programming (e.g., Mixed Integer Programming, MIP) or Constrained Programming (CP) to model and solve scheduling problems with complex constraints in order to find mathematically optimal solutions.

[0005] For example, patent CN113610233A discloses a "Flexible Job Shop Scheduling Method Based on Improved Genetic Algorithm". The focus of this patent's technical solution is to improve the algorithm's performance and convergence speed in solving the static Flexible Job Shop Scheduling Problem (FJSP) by specifically designing the encoding method, crossover, and mutation operators of the genetic algorithm. However, this patent's technical solution is essentially still an offline optimization algorithm oriented towards static problem instances. It does not disclose a system architecture capable of responding to real-time disturbances in the production site, nor does it involve a technology for deeply integrating configurable business rules with the optimization solution process. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a real-time scheduling method and system for PCB production based on hybrid intelligent decision-making, which solves the technical problems in the prior art.

[0007] In a first aspect, the present invention provides the following technical solution: a real-time scheduling method for PCB production based on hybrid intelligent decision-making, comprising: Obtain PCB production information and construct a corresponding digital twin for the production information; Based on the state identification of the digital twin, explicit and implicit process tasks in the production work order are identified. Obtain the current business strategy and load the hard constraint rule set and soft constraint rule set based on the current business strategy; By combining the explicit process tasks, the implicit process tasks, the hard constraint rule set, and the soft constraint rule set, a scheduling problem instance is obtained; The scheduling problem instance is transformed into an objective function, and the objective function is solved by multi-objective optimization to output an initial scheduling plan; The initial scheduling plan is monitored in real time and dynamically incrementally rescheduled in a loop to complete the real-time scheduling of PCB production.

[0008] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention realizes a paradigm shift from 'offline batch processing' to 'online real-time decision-making', revolutionizing production agility. Existing technologies are generally offline and static, while this invention, through event-triggered real-time data perception and incremental planning technology, shortens the decision-making cycle from 'hours' to 'minutes'. When machine failures or emergency orders occur, the system can automatically generate globally coordinated response plans, transforming production plans from rigid instructions into dynamically adaptable 'living' entities, fundamentally improving the flexibility and market responsiveness of the production system. This invention significantly improves the quality and on-site feasibility of scheduling results: Through a hybrid rule engine, this invention makes implicit expert knowledge explicit into a series of quantifiable soft constraints and deeply embeds them into the scoring function of the optimization algorithm. This allows the algorithm to "understand" the subtle preferences of the business at each step of the optimization process (such as minimizing line changes and balancing equipment load). The generated solutions are not only mathematically superior, but also more reasonable in production practice, effectively improving the overall OEE (Overall Equipment Effectiveness) and reducing production costs. This invention greatly enhances the flexibility and maintainability of the system: existing APS systems have fixed rules, and changes to business processes require expensive and time-consuming secondary development. This invention adopts a design that decouples business rules from core algorithms and parameterizes and configures rules and weights, allowing business personnel (such as production managers) to adjust the weights of different optimization objectives (such as delivery time and cost) through the interface, and quickly respond to changes in corporate strategy. This invention comprehensively improves the efficiency of scheduling and the level of intelligence in enterprise operations: It employs an advanced metaheuristic optimization algorithm, capable of finding high-quality solutions within an acceptable timeframe when dealing with large-scale, highly complex PCB scheduling problems. More importantly, the entire closed-loop process of "perception-decision-execution-feedback" is highly automated, completely freeing schedulers from tedious, repetitive, and error-prone manual planning and adjustment work, transforming them into "commanders" on the production floor, and significantly enhancing the overall intelligence level of production operations.

[0009] Preferably, the step of obtaining PCB production information and constructing a corresponding digital twin for the production information includes: Acquire work order data, resource data, material and tool data, and real-time events from PCB production to obtain production information; A data structure is constructed for the consumable tools in the production information, the data structure including static IDs and dynamically updated consumption data, to obtain a micro-digital twin model; A macro-level digital twin model is constructed for other data in the production information. The micro-level digital twin model and the macro-level digital twin model are then combined to obtain the corresponding digital twin.

[0010] Preferably, the step of identifying explicit and implicit process tasks in the production work order based on the state recognition of the digital twin includes: The work orders in the digital twin are broken down into several process tasks arranged according to the process route; The available resource information, available tool information, and order information during the scheduling process are used as decision information. The decision information is then associated with the process tasks to obtain explicit process tasks. Based on the process task, a process requirement analysis is performed. If there are cases where the process requirement analysis is not satisfactory, a hidden task that meets the process requirement analysis is created to obtain the hidden process task.

[0011] Preferably, in the step of obtaining the current business strategy and loading the hard constraint rule set and the soft constraint rule set based on the current business strategy, the hard constraint rule set includes machine capacity constraints, process sequence constraints, resource exclusivity constraints, and tool availability constraints, and the soft constraint rule set includes delivery time optimization rules, line changeover minimization rules, load balancing rules, and machine optimization rules.

[0012] Preferably, the step of converting the scheduling problem instance into an objective function and performing multi-objective optimization on the objective function to output an initial scheduling plan includes: Transform the scheduling problem instance into an objective function. : ; ; ; ; ; ; In the formula, These are respectively the order set, the machine set, and the process task set. , , , These represent the penalty items corresponding to delivery delay penalties, changeover time penalties, tool-task mismatch penalties, and other soft constraint rules, respectively. , Orders Start time, required delivery date, For the machine On the task Execute the task immediately upon completion. Required changeover time These are the first, second, and third binary variables, respectively. Representing tasks Start time, task End time, task End time, For orders Priority weights, For a set of priority relations, These are the first soft constraint weight, the second soft constraint weight, and the third soft constraint weight, respectively. Add operators and use an adaptive learning mechanism to solve the objective function using the operators to obtain an initial scheduling plan.

[0013] Preferably, in the step of adding operators and using an adaptive learning mechanism to solve the objective function using the operators to obtain the initial scheduling plan, the operators include batch merging move operators, tool-task collaborative move operators, and high- and low-precision task swap move operators; The batch merging move operator is used to process similar tasks in batches and reduce tool-changing tasks caused by tool changes, thereby reducing the value of the tool-changing time penalty in the objective function; The tool-task cooperative movement operator is used to reduce the value of the tool-change time penalty in the objective function by controlling the machine to use existing tools and avoiding the generation of tool-change tasks; The high- and low-precision task swapping operator is used to identify task resource mismatches in the current solution of the objective function and swap the resource allocation of tasks according to the resource mismatches, so as to reduce the value of the tool-task mismatch penalty in the objective function.

[0014] Preferably, the step of performing real-time monitoring and dynamic incremental rescheduling of the initial scheduling plan to complete the real-time scheduling of PCB production includes: Real-time monitoring of changes in the production site status and external event flow; and determination of whether rescheduling conditions are triggered based on the changes in status and external event flow. If the rescheduling conditions are not triggered, production and processing will proceed according to the initial scheduling plan. If the rescheduling condition is triggered, the digital twin is updated and all completed process tasks within a rolling freeze time window are locked, and the state change and the external event stream are updated in the current scheduling problem instance. The updated scheduling problem instance is transformed into an update objective function, and the first soft constraint weight, the second soft constraint weight, and the third soft constraint weight in the update objective function are updated as follows: ; ; In the formula, For the first The first, second, or third soft constraint weights after the second rescheduling iteration For the updated first soft constraint weight, second soft constraint weight, or third soft constraint weight, They represent the first sequence The current error of the second rescheduling iteration. These are the proportional action weight, integral action weight, and differential action weight, respectively. Update the weights of the operators: ; In the formula, For learning rate, For the weights of the updated operator, The weights of the operators before the update. This represents the objective function value of the operator this week. Based on the updated first soft constraint weights, second soft constraint weights, third soft constraint weights, and updated operator weights, an adaptive learning mechanism is used to solve the update objective function to obtain the updated scheduling plan and publish the updated scheduling plan.

[0015] Secondly, the present invention provides the following technical solution: a real-time scheduling system for PCB production based on hybrid intelligent decision-making, the system comprising: The acquisition module is used to acquire PCB production information and construct a corresponding digital twin for the production information; The task module is used to identify explicit and implicit process tasks in the production work order based on the state of the digital twin. The constraint module is used to obtain the current business strategy and load hard constraint rule sets and soft constraint rule sets based on the current business strategy; The instance module is used to synthesize the explicit process tasks, the implicit process tasks, the hard constraint rule set, and the soft constraint rule set to obtain scheduling problem instances; The solution module is used to convert the scheduling problem instance into an objective function, perform multi-objective optimization on the objective function, and output an initial scheduling plan. The loop module is used to monitor the initial scheduling plan in real time and perform dynamic incremental rescheduling loops to complete the real-time scheduling of PCB production.

[0016] Thirdly, the present invention provides the following technical solution: a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described real-time scheduling method for PCB production based on hybrid intelligent decision-making.

[0017] Fourthly, the present invention provides the following technical solution: a storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the above-described real-time scheduling method for PCB production based on hybrid intelligent decision-making. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart of the real-time scheduling method for PCB production based on hybrid intelligent decision-making provided in Embodiment 1 of the present invention; Figure 2 This is a structural block diagram of the PCB production real-time scheduling system based on hybrid intelligent decision-making provided in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the hardware structure of a computer provided for another embodiment of the present invention.

[0020] The embodiments of the present invention will be further described below with reference to the accompanying drawings. Detailed Implementation

[0021] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain embodiments of the present invention, and should not be construed as limiting the present invention.

[0022] Example 1 In Embodiment 1 of the present invention, as Figure 1 As shown, a real-time scheduling method for PCB production based on hybrid intelligent decision-making includes: S1. Obtain PCB production information and construct a corresponding digital twin for the production information; Specifically, the main sources of production information here are: MES / ERP / WMS systems provide work order data (process routes, delivery dates, priorities), material inventory, etc. Equipment Internet of Things (IoT) platform: provides real-time status of machines (running, idle, faulty), capacity parameters, historical processing yield, etc. Tool / Quality Management System: Provides information on the availability status, remaining useful life (RUL), and quality anomaly events of tools such as drill bits; Meanwhile, the constraints mentioned later also come from the aforementioned platform and system.

[0023] Step S1 includes: S11. Obtain PCB production work order data, resource data, material and tool data, and real-time events to obtain production information; The work order data includes work order number, product part number, quantity, process route (including alternative processes), customer priority, promised delivery date, and technical requirements (such as board thickness and copper thickness); resource data includes machine ID, machine group, capability parameters (such as drilling speed, accuracy level, and lamination temperature range), current status (running, idle, faulty, planned maintenance, changeover), current processing task, assigned task queue, estimated available time, and historical processing yield; material and tool data includes inventory, location, and estimated arrival time of materials such as PCB substrates, chemicals, and drill bits; and the availability, lifespan, and location of tools such as drill bits and films; real-time events include machine fault alarms, quality control hold (QC Hold), emergency order insertion requests, order cancellation or modification, material delivery arrival signals, and operator login / logout. This is implemented using a service-oriented architecture (SOA) through RESTful API interfaces, database change data capture (CDC), and distributed message queues (such as Kafka). The combination of technologies such as RabbitMQ ensures that any changes in the production site can be detected at the second or sub-second level, providing a solid real-time data foundation for system decision-making.

[0024] S12. Construct a data structure for consumable tools in the production information, the data structure including static IDs and dynamically updated consumption data, to obtain a micro-digital twin model. Specifically, taking drill bits as an example of consumable tools, a digital twin object is created for each individual drill bit. Its data structure includes a static ID and dynamically updated consumable attributes, such as the number of holes drilled (hitCount) and the Remaining Useful Life (RUL). RUL is calculated in real-time based on the number of holes drilled, the cumulative depth, and the hardness coefficient of the processed material, using a preset wear model. The Remaining Useful Life is not static but has online self-learning and correction capabilities. The system continuously collects event data related to actual drill bit failure or deterioration in processing quality (e.g., "excessive burrs" events reported by the quality management system, or "drill bit chipping" events entered by the operator). When the status of a drill bit twin is updated to "end of life," the system compares the difference between its "predicted life" and "actual life." This difference serves as an error signal, which is then used by an adaptive algorithm (e.g., simple least squares or more complex Bayesian updates) to fine-tune the core parameters (such as the hardness wear coefficient of different materials) of this type of drill bit in the wear model. In this way, digital twin models can continuously learn from historical data, making their predictions increasingly closer to physical reality, thus achieving a leap from "passive mapping" to "active evolution".

[0025] S13. Construct a macroscopic digital twin model for other data in the production information, and combine the microscopic digital twin model and the macroscopic digital twin model to obtain the corresponding digital twin; Specifically, through the above processing, relational data and event data are instantiated into a set of interconnected, high-fidelity standardized object models that include macro-level equipment, micro-level tools, and implicit preparation activities. This provides a solid data foundation for achieving accurate and pragmatic scheduling optimization. In practice, the twin's status is updated in real time. Once a status event is received, such as a drilling task completion event, the twin automatically updates its hitCount and RUL values ​​based on information such as the number of holes processed carried in the event. When RUL is lower than the warning threshold, the twin will proactively issue a "tool life warning event" to provide a basis for subsequent scheduling decisions.

[0026] S2. Based on the state identification of the digital twin, identify explicit and implicit process tasks in the production work order; Step S2 includes: S21. The work order in the digital twin is split into several process tasks arranged according to the process route; Specifically, the process tasks here are the planning entities, which specifically refer to the objects that need to be scheduled. Each work order is broken down into a series of process tasks ordered according to the process route. The task object not only includes its basic processing time, but also includes attributes such as setupDuration (changeover time, which may depend on the preceding task) and requiredToolType (required tool type). Its core planning variables are the assigned machine and the start time.

[0027] S22. Use available resource information, available tool information, and order information during the scheduling process as decision information, and associate the decision information with the process task to obtain explicit process task; Specifically, the decision information here refers to information that remains constant in the scheduling process or serves as the basis for decision-making. For example, a Machine object has properties including a SkillSet (a set of capability tags) and a MaintenanceSchedule (a maintenance schedule time window). An Order object has properties including PriorityLevel and DueDate.

[0028] S23. Analyze the process requirements based on the process task. If there are cases where the process requirements analysis is not qualified, create a hidden task that meets the process requirements analysis to obtain the hidden process task. In this application, when parsing a work order into a task, the system analyzes the task's technological requirements (such as the required drill bit type). If the required drill bit does not match the one currently installed on the machine, or if the current drill bit's RUL cannot meet the task's accuracy requirements, the system will automatically inject an independent "ToolChangeTask" before the task. This ToolChangeTask has independent attributes such as duration and resource requirements, and participates in the subsequent optimization solution as an independent planning entity. Specifically, in addition to the "tool changing task," the implicit process tasks may also include, but are not limited to: necessary machine preheating tasks, specific chemical solution replacement and adjustment tasks, or automated cleaning tasks required between processing two significantly different products. The system can automatically identify and inject these necessary preparatory activities to ensure production quality and continuity based on a preset process knowledge base and the attribute differences between preceding and following process tasks, thereby ensuring the integrity and actual executability of the scheduling plan.

[0029] S3. Obtain the current business strategy and load the hard constraint rule set and soft constraint rule set based on the current business strategy; Specifically, the hard constraint rule set includes machine capacity constraints, process sequence constraints, resource exclusivity constraints, and tool availability constraints, while the soft constraint rule set includes delivery time optimization rules, line changeover minimization rules, load balancing rules, and machine optimization rules. Hard constraints are rules that must be strictly followed; violating any one of them will render the scheduling plan invalid. Examples include machine capacity constraints, process sequence constraints, resource exclusivity constraints, and tool availability constraints.

[0030] Soft constraint rules are rules that are desired to be met as much as possible, but are not mandatory. Each violation of a soft constraint incurs a penalty score based on its importance. The optimization goal is to minimize the total penalty score. Examples include: delivery time optimization rules, line swapping minimization rules, load balancing rules, and machine selection rules. Meanwhile, in this application, the above rules implement dynamic soft constraints oriented towards production scenarios, and their weights can be adaptively adjusted according to the real-time situation in the workshop. These weights are respectively the rule logic and the scenario adaptive weights: Rule logic: When a drill bit's remaining effective life (RUL) is nearing the end of its life, if the system assigns it to a drilling task for a low-precision or non-critical product, the scheduling scheme will receive bonus points; conversely, if it is assigned to a high-precision task, penalty points will be generated. Context-Adaptive Weighting: The penalty weight of this rule is not fixed. When the number of high-precision orders backlogged in the system exceeds a threshold, the system will automatically increase the penalty weight of this constraint to force high-precision tasks to use new tools in order to prioritize quality. Conversely, when production pressure is low, its weight will be appropriately reduced to encourage the use of old tools and save costs.

[0031] S4. Combine the explicit process tasks, the implicit process tasks, the hard constraint rule set, and the soft constraint rule set to obtain a scheduling problem instance; Specifically, in this step, based on process requirements and the state of the tool twin, necessary implicit process tasks will be automatically identified and injected as independent planning entities. Finally, all entities to be planned, including explicit and implicit process tasks, as well as all resources and constraints, will be transformed into a complete scheduling problem instance.

[0032] S5. Convert the scheduling problem instance into an objective function, perform multi-objective optimization on the objective function, and output the initial scheduling plan; Step S5 includes: S51. Convert the scheduling problem instance into an objective function. : ; ; ; ; ; ; In the formula, These are respectively the order set, the machine set, and the process task set. , , , These represent the penalty items corresponding to delivery delay penalties, changeover time penalties, tool-task mismatch penalties, and other soft constraint rules, respectively. , Orders Start time, required delivery date, For the machine On the task Execute the task immediately upon completion. Required changeover time These are the first, second, and third binary variables, respectively. Representing tasks Start time, task End time, task End time, For orders Priority weights, For a set of priority relations, These are the first soft constraint weight, the second soft constraint weight, and the third soft constraint weight, respectively. Among them, for the first, second, and third binary variables, when the process task Assigned to machine During execution, The value of is 1 if it is not 0 otherwise; the same applies to the second binary variable, when the process task Assigned to machine During execution, The value is 1 if the task is not specified, and 0 otherwise. and process tasks At the machine When the components are arranged to be processed in an adjacent sequence, the value is 1; otherwise, it is 0. This is used to determine whether a changeover time will occur. Specifically, the penalty terms corresponding to other soft constraint rules here can be those for load balancing, machine optimization, etc. Furthermore, in practice, the objective function also requires tool availability constraints, i.e., the availability of tools used during the execution of a given task. It must be available.

[0033] S52. Add operators and use an adaptive learning mechanism to solve the objective function using the operators to obtain the initial scheduling plan; The operators include batch merging move operators, tool-task collaborative move operators, and high- and low-precision task swapping move operators; Specifically, the actual solution process is divided into two steps. The first step is the construction of a high-quality initial solution. The optimization does not start with a random solution, but rather first runs a domain-customized constructivist heuristic algorithm. This algorithm follows the principles of "bottleneck priority" and "critical chain," prioritizing tasks that are expected to be the most time-consuming or resource-intensive (such as high-precision machines required for special materials) and placing them in their optimal positions, thereby quickly generating an initial solution with a good framework. The second step is adaptive large neighborhood search iterative optimization, which is the core iterative optimization stage. Its framework includes two types of core operators: Destruction operators, responsible for removing a portion of the tasks from the current solution to create "repair" space; and Repair operators, responsible for re-inserting the removed tasks into the solution in a better manner. The aforementioned batch merging move operator, tool-task collaborative move operator, and high-low precision task swap move operator can all be used as destruction and repair operators. During the solution process, the frequency and magnitude of successful solution improvement for each "destruction-repair" operator combination in recent iterations are statistically analyzed in real time. The probability of selecting each operator is dynamically adjusted based on the statistical results, enabling the algorithm to learn and automatically favor operators that are more effective in the current solution stage, thereby greatly improving search efficiency.

[0034] The batch merging move operator is used to process similar tasks in batches and reduce tool-changing tasks caused by tool changes, thereby reducing the value of the tool-changing time penalty in the objective function; Among them, the batch merging move operator is an advanced operator that identifies tasks that use the same tools (such as the same type of drill bit) but are scheduled at different times or on different machines. It "disrupts" these tasks and then, in the "repair" phase, it continuously schedules them on the same machine. Its optimization objective is to minimize the "tool change tasks" caused by changing tools by processing similar tasks in "batch" and reduce the value of the tool change time penalty in the objective function.

[0035] The tool-task cooperative movement operator is used to reduce the value of the tool-change time penalty in the objective function by controlling the machine to use existing tools and avoiding the generation of tool-change tasks; The tool-task cooperative movement operator is an operator with heuristic logic, which is used as a repair operator in this application. When it needs to find a new insertion position for a task, it will give priority to those machines that have already installed the required tools. Its optimization objective is similar to "batch merging". It aims to avoid tool changes by "making do" with the existing tools of the machine, so as to reduce the value of the changeover time penalty in the objective function.

[0036] The high- and low-precision task swapping operator is used to identify task resource mismatches in the current solution of the objective function and swap the resource allocation of tasks according to the resource mismatches, so as to reduce the value of the tool-task mismatch penalty in the objective function. Among them, the high-precision and low-precision task swapping operator is a highly targeted composite operator. It actively searches for a specific "resource mismatch" in the current solution: a high-precision task uses an old tool at the end of its lifespan, while a low-precision task occupies a brand-new tool. After finding such a "mismatch", it swaps the machine or tool allocation of the two tasks. Its optimization objective is to ensure that "good tools are used where they are most needed". It directly improves the score of the soft constraint rule of "maximizing the value of tools at the end of their lifespan" by ensuring that "good tools are used where they are most needed". This reduces the value of the tool-task mismatch penalty in the objective function and achieves dual optimization of quality and cost. Specifically, in order to realize the high-precision and low-precision task swapping movement operator, the system will automatically mark each process task with the attribute of "high precision" or "normal precision" in advance according to the technical requirements in the acquired work order data (e.g., specific drilling tolerance, line accuracy level, or customer-specified key process). This attribute serves as an objective basis for accurately identifying resource mismatch during the solution process, thereby ensuring the effectiveness of operator execution.

[0037] In summary, at each step of the iterative optimization, the algorithm attempts to apply these operators to generate new solutions. If a new solution reduces the total score (total penalty score) of the objective function Z, then this "move" is a successful optimization. Through tens of thousands of such attempts, the entire scheduling scheme will gradually approach the optimum.

[0038] S6. Perform real-time monitoring and dynamic incremental rescheduling cycles on the initial scheduling plan to complete the real-time scheduling of PCB production.

[0039] Step S6 includes: S61. Monitor the status changes and external event flow of the production site in real time, and determine whether the rescheduling conditions are triggered based on the status changes and external event flow. Specifically, the rescheduling conditions here can be periodically triggered, such as performing rolling optimization every 15 minutes, or they can be event-triggered, such as receiving a machine fault alarm or inserting a new high-priority work order.

[0040] S62. If the rescheduling condition is not triggered, production and processing shall be carried out according to the initial scheduling plan. S63. If the rescheduling condition is triggered, update the digital twin and lock all completed process tasks within a rolling freeze time window, and update the state change and the external event stream to the current scheduling problem instance. Specifically, if triggered, the system immediately updates the digital twin model and automatically marks all planned tasks within a preset, rolling freeze time window as 'locked' to ensure production stability. At the same time, new changes (such as faulty machines becoming unavailable within a specific time period or newly added work orders) are updated to the current scheduling problem instance.

[0041] S64. Convert the updated scheduling problem instance into an update objective function, and update the first soft constraint weight, the second soft constraint weight, and the third soft constraint weight in the update objective function: ; ; In the formula, For the first The first, second, or third soft constraint weights after the second rescheduling iteration For the updated first soft constraint weight, second soft constraint weight, or third soft constraint weight, They represent the first sequence The current error of the second rescheduling iteration. These are the proportional action weight, integral action weight, and differential action weight, respectively. Specifically, the update process here is iterative. Each reordering update updates the current result based on the previous reordering. In this step, a PID algorithm is used to update the weights of the soft constraints. In practice, a key performance indicator (KPI)-based update method is introduced. First, a target value (e.g., 98%) is set for the core business objective (e.g., "on-time delivery rate of high-priority orders"). After each scheduling, the achievement of this KPI is predicted based on the scheduling results. The strategy controller continuously compares the "predicted KPI value" with the "target KPI value." If the predicted value is lower than the target value, the controller will automatically and smoothly increase the weights of soft constraints directly related to this KPI (e.g., "delivery date optimization rules") in the input of the next reordering using a preset PID controller. Conversely, if the predicted value is much higher than the target, the weights can be appropriately reduced to free up optimization space for other optimization objectives (e.g., cost). This mechanism transforms the adjustment of business strategies from discrete, threshold-based judgments to continuous, feedback-based, and automated closed-loop control. Here, the current error of the nth rescheduling iteration. Precisely defined as the normalized deviation between the predicted and target KPI values ​​of this scheduling plan, it can be expressed as: =(KPItarget-KPIpredicted,n) / KPItarget. Where KPIpredicted,n is the predicted on-time delivery rate and other indicators calculated based on the current scheduling results, which is the predicted KPI value, and KPItarget is the target KPI value.

[0042] The calculation process for the predicted KPI values ​​here is as follows: Iterate through all orders marked as "high priority" in the scheduling plan. For each order, find the completion time of its last task and compare the completion time with the required delivery date of the order. If the completion time is less than or equal to the required delivery date, the order is judged as on time. Finally, the predicted KPI value is: (number of high priority orders completed on time / total number of high priority orders) × 100%.

[0043] S65. Update the weights of the operators: ; In the formula, For learning rate, For the weights of the updated operator, The weights of the operators before the update. This represents the objective function value of the operator this week. Specifically, in the initial stage, an initial weight is assigned to each operator (or each "destruction-repair" operator combination). At the beginning, all operators have an equal probability of being selected. In each iteration of optimization, a pair of "destruction-repair" operators needs to be selected to modify the current solution. This selection process is based on probability and usually uses methods such as "roulette wheel". The higher the score of the operator, the greater the probability of it being selected. After the operator combination is applied, a new scheduling solution is generated. The quality of this new solution (i.e., the total penalty score) is evaluated. Based on the evaluation results, the scores of the selected operators are updated in real time. If the new solution is significantly better than the current optimal solution, the operator combination receives a high score reward; if the new solution is only slightly better, it receives a lower reward. Under certain algorithmic mechanisms (such as simulated annealing), a poor solution may be accepted to escape local optima. In this case, the operator score may remain unchanged or increase slightly; if the new solution does not improve and is rejected, the operator receives no score.

[0044] The learning process is carried out in "segments" (for example, every 100 iterations constitute a segment). After a segment ends, the system updates the selection weights of each operator for the next segment based on the accumulated scores of each operator during the segment. The update formula is as shown above.

[0045] S66. Based on the updated first soft constraint weight, second soft constraint weight, third soft constraint weight, and updated operator weight, an adaptive learning mechanism is used to solve the update objective function to obtain the updated scheduling plan and publish the updated scheduling plan.

[0046] The PCB production real-time scheduling method based on hybrid intelligent decision-making provided in Embodiment 1 of this invention has the following advantages compared to the prior art: This invention realizes a paradigm shift from 'offline batch processing' to 'online real-time decision-making', revolutionizing production agility. The prior art is generally offline and static, while this invention, through event-triggered real-time data perception and incremental planning technology, shortens the decision cycle from 'hours' to 'minutes'. When machine failure or emergency order insertion occurs, the system can automatically generate a globally coordinated response plan, transforming the production plan from a rigid instruction into a dynamically adaptable 'living' entity, fundamentally improving the flexibility and market responsiveness of the production system. This invention significantly improves the quality and on-site feasibility of scheduling results: Through a hybrid rule engine, this invention makes implicit expert knowledge explicit into a series of quantifiable soft constraints and deeply embeds them into the scoring function of the optimization algorithm. This allows the algorithm to "understand" the subtle preferences of the business at each step of the optimization process (such as minimizing line changes and balancing equipment load). The generated solutions are not only mathematically superior, but also more reasonable in production practice, effectively improving the overall OEE (Overall Equipment Effectiveness) and reducing production costs. This invention greatly enhances the flexibility and maintainability of the system: existing APS systems have fixed rules, and changes to business processes require expensive and time-consuming secondary development. This invention adopts a design that decouples business rules from core algorithms and parameterizes and configures rules and weights, allowing business personnel (such as production managers) to adjust the weights of different optimization objectives (such as delivery time and cost) through the interface, and quickly respond to changes in corporate strategy. This invention comprehensively improves the efficiency of scheduling and the level of intelligence in enterprise operations: It employs an advanced metaheuristic optimization algorithm, capable of finding high-quality solutions within an acceptable timeframe when dealing with large-scale, highly complex PCB scheduling problems. More importantly, the entire closed-loop process of "perception-decision-execution-feedback" is highly automated, completely freeing schedulers from tedious, repetitive, and error-prone manual planning and adjustment work, transforming them into "commanders" on the production floor, and significantly enhancing the overall intelligence level of production operations.

[0047] Example 2 like Figure 2 As shown, in Embodiment 2 of the present invention, a real-time scheduling system for PCB production based on hybrid intelligent decision-making is provided. The system includes: Module 1 is used to acquire PCB production information and construct a corresponding digital twin for the production information; Task module 2 is used to identify explicit and implicit process tasks in production work orders based on the state of the digital twin. Constraint module 3 is used to obtain the current business strategy and load hard constraint rule sets and soft constraint rule sets based on the current business strategy; Instance module 4 is used to integrate the explicit process tasks, the implicit process tasks, the hard constraint rule set, and the soft constraint rule set to obtain scheduling problem instances; Solving module 5 is used to convert the scheduling problem instance into an objective function, perform multi-objective optimization on the objective function, and output an initial scheduling plan. The loop module 6 is used to monitor the initial scheduling plan in real time and perform dynamic incremental rescheduling loop to complete the real-time scheduling of PCB production. The acquisition module 1 includes: The information submodule is used to acquire work order data, resource data, material and tool data, and real-time events from PCB production to obtain production information. The structural submodule is used to construct a data structure for consumable tools in the production information. The data structure includes static IDs and dynamically updated consumption data to obtain a micro-digital twin model. The twin submodule is used to construct a macro-level digital twin model for other data in the production information, and to combine the micro-level digital twin model and the macro-level digital twin model to obtain the corresponding digital twin.

[0048] The task module 2 includes: The splitting submodule is used to split the work order in the digital twin into several process tasks arranged according to the process route; The association submodule is used to use available resource information, available tool information, and order information during the scheduling process as decision information, and associate the decision information with the process task to obtain explicit process task; The analysis submodule is used to perform process requirement analysis based on the process task. If there is a case where the process requirement analysis is not qualified, a hidden task that meets the process requirement analysis is created to obtain the hidden process task.

[0049] The solution module 5 includes: The function submodule is used to convert the scheduling problem instance into an objective function. : ; ; ; ; ; ; In the formula, These are respectively the order set, the machine set, and the process task set. , , , These represent the penalty items corresponding to delivery delay penalties, changeover time penalties, tool-task mismatch penalties, and other soft constraint rules, respectively. , Orders Start time, required delivery date, For the machine On the task Execute the task immediately upon completion. Required changeover time These are the first, second, and third binary variables, respectively. Representing tasks Start time, task End time, task End time, For orders Priority weights, For a set of priority relations, These are the first soft constraint weight, the second soft constraint weight, and the third soft constraint weight, respectively. The solver submodule is used to add operators and use an adaptive learning mechanism to solve the objective function using the operators to obtain the initial scheduling plan.

[0050] The loop module 6 includes: The monitoring submodule is used to monitor the status changes and external event flow of the production site in real time, and determine whether the rescheduling conditions are triggered based on the status changes and external event flow. The condition submodule is used to carry out production and processing according to the initial schedule if the rescheduling condition is not triggered. The locking submodule is used to update the digital twin and lock all completed process tasks within a rolling freeze time window if a rescheduling condition is triggered, and update the state change and the external event stream to the current scheduling problem instance. The first update submodule is used to convert the updated scheduling problem instance into an update objective function, and update the first soft constraint weight, the second soft constraint weight, and the third soft constraint weight in the update objective function: ; ; In the formula, For the first The first, second, or third soft constraint weights after the second rescheduling iteration For the updated first soft constraint weight, second soft constraint weight, or third soft constraint weight, They represent the first sequence The current error of the second rescheduling iteration. These are the proportional action weight, integral action weight, and differential action weight, respectively. The second update submodule is used to update the weights of the operators: ; In the formula, For learning rate, For the weights of the updated operator, The weights of the operators before the update. This represents the objective function value of the operator this week. The third update submodule is used to solve the update objective function using an adaptive learning mechanism based on the updated first soft constraint weights, second soft constraint weights, third soft constraint weights, and updated operator weights, so as to obtain the updated scheduling plan and publish the updated scheduling plan.

[0051] In other embodiments of the present invention, the present invention provides the following technical solution: a computer, including a memory 102, a processor 101, and a computer program stored on the memory 102 and executable on the processor 101, wherein the processor 101 executes the computer program to implement the PCB production real-time scheduling method based on hybrid intelligent decision-making as described above.

[0052] Specifically, the processor 101 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of the present invention.

[0053] The memory 102 may include a large-capacity memory for data or instructions. For example, and not limitingly, the memory 102 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 102 may include removable or non-removable (or fixed) media. Where appropriate, the memory 102 may be internal or external to a data processing device. In a particular embodiment, the memory 102 is non-volatile memory. In a particular embodiment, the memory 102 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random Access Memory (FPMDRAM), Extended Data Out Dynamic Random Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0054] The memory 102 can be used to store or cache various data files that need to be processed and / or used for communication, as well as possible computer program instructions executed by the processor 101.

[0055] The processor 101 reads and executes the computer program instructions stored in the memory 102 to implement the above-mentioned real-time scheduling method for PCB production based on hybrid intelligent decision-making.

[0056] In some embodiments, the computer may further include a communication interface 103 and a bus 100. For example, Figure 3 As shown, the processor 101, memory 102, and communication interface 103 are connected through bus 100 and complete communication with each other.

[0057] The communication interface 103 is used to enable communication between the various modules, devices, units, and / or equipment in the embodiments of the present invention. The communication interface 103 can also enable data communication with other components such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations.

[0058] Bus 100 includes hardware, software, or both, that couples components of a computer device together. Bus 100 includes, but is not limited to, at least one of the following: data bus, address bus, control bus, expansion bus, and local bus. For example, and not as a limitation, bus 100 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 100 may include one or more buses. Although specific buses are described and illustrated in the embodiments of the present invention, the present invention is contemplated by any suitable bus or interconnect.

[0059] The computer can execute the PCB production real-time scheduling method based on hybrid intelligent decision-making of the present invention based on the PCB production real-time scheduling system obtained from the hybrid intelligent decision-making system, thereby realizing PCB production real-time scheduling based on hybrid intelligent decision-making.

[0060] In some further embodiments of the present invention, in conjunction with the above-described real-time scheduling method for PCB production based on hybrid intelligent decision-making, the present invention provides the following technical solution: a storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the above-described real-time scheduling method for PCB production based on hybrid intelligent decision-making.

[0061] Those skilled in the art will understand that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0062] More specific examples of readable media (a non-exhaustive list) include: electrical connections (electronic devices) with one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0063] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0064] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0065] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A real-time scheduling method for PCB production based on hybrid intelligent decision-making, characterized in that, include: Obtain PCB production information and construct a corresponding digital twin for the production information; Based on the state identification of the digital twin, explicit and implicit process tasks in the production work order are identified. Obtain the current business strategy and load the hard constraint rule set and soft constraint rule set based on the current business strategy; By combining the explicit process tasks, the implicit process tasks, the hard constraint rule set, and the soft constraint rule set, a scheduling problem instance is obtained; The scheduling problem instance is transformed into an objective function, and the objective function is solved by multi-objective optimization to output an initial scheduling plan; The initial scheduling plan is monitored in real time and dynamically incrementally rescheduled in a loop to complete the real-time scheduling of PCB production.

2. The PCB production real-time scheduling method based on hybrid intelligent decision-making according to claim 1, characterized in that, The steps of obtaining PCB production information and constructing a corresponding digital twin for the production information include: Acquire work order data, resource data, material and tool data, and real-time events from PCB production to obtain production information; A data structure is constructed for the consumable tools in the production information, the data structure including static IDs and dynamically updated consumption data, to obtain a micro-digital twin model; A macro-level digital twin model is constructed for other data in the production information. The micro-level digital twin model and the macro-level digital twin model are then combined to obtain the corresponding digital twin.

3. The PCB production real-time scheduling method based on hybrid intelligent decision-making according to claim 1, characterized in that, The steps for identifying explicit and implicit process tasks in a production work order based on the state recognition of the digital twin include: The work orders in the digital twin are broken down into several process tasks arranged according to the process route; The available resource information, available tool information, and order information during the scheduling process are used as decision information. The decision information is then associated with the process tasks to obtain explicit process tasks. Based on the process task, a process requirement analysis is performed. If there are cases where the process requirement analysis is not satisfactory, a hidden task that meets the process requirement analysis is created to obtain the hidden process task.

4. The PCB production real-time scheduling method based on hybrid intelligent decision-making according to claim 1, characterized in that, In the step of obtaining the current business strategy and loading the hard constraint rule set and soft constraint rule set based on the current business strategy, the hard constraint rule set includes machine capacity constraints, process sequence constraints, resource exclusivity constraints, and tool availability constraints, and the soft constraint rule set includes delivery time optimization rules, line changeover minimization rules, load balancing rules, and machine optimization rules.

5. The PCB production real-time scheduling method based on hybrid intelligent decision-making according to claim 1, characterized in that, The steps of converting the scheduling problem instance into an objective function and performing multi-objective optimization on the objective function to output an initial scheduling plan include: Transform the scheduling problem instance into an objective function. : ； ； ； ； ； ； In the formula, These are respectively the order set, the machine set, and the process task set. , , , These represent the penalty items corresponding to delivery delay penalties, changeover time penalties, tool-task mismatch penalties, and other soft constraint rules, respectively. , Orders Start time, required delivery date, For the machine On the task Execute the task immediately upon completion. Required changeover time These are the first, second, and third binary variables, respectively. Representing tasks Start time, task End time, task End time, For orders Priority weights, For a set of priority relations, These are the first soft constraint weight, the second soft constraint weight, and the third soft constraint weight, respectively. Add operators and use an adaptive learning mechanism to solve the objective function using the operators to obtain an initial scheduling plan.

6. The PCB production real-time scheduling method based on hybrid intelligent decision-making according to claim 5, characterized in that, In the step of adding operators and using an adaptive learning mechanism to solve the objective function using the operators to obtain the initial scheduling plan, the operators include batch merging move operators, tool-task collaborative move operators, and high- and low-precision task swap move operators; The batch merging move operator is used to process similar tasks in batches and reduce tool-changing tasks caused by tool changes, thereby reducing the value of the tool-changing time penalty in the objective function; The tool-task cooperative movement operator is used to reduce the value of the tool-change time penalty in the objective function by controlling the machine to use existing tools and avoiding the generation of tool-change tasks; The high- and low-precision task swapping operator is used to identify task resource mismatches in the current solution of the objective function and swap the resource allocation of tasks according to the resource mismatches, so as to reduce the value of the tool-task mismatch penalty in the objective function.

7. The PCB production real-time scheduling method based on hybrid intelligent decision-making according to claim 1, characterized in that, The steps of real-time monitoring and dynamic incremental rescheduling of the initial scheduling plan to complete the real-time scheduling of PCB production include: Real-time monitoring of changes in the production site status and external event flow; and determination of whether rescheduling conditions are triggered based on the changes in status and external event flow. If the rescheduling conditions are not triggered, production and processing will proceed according to the initial scheduling plan. If the rescheduling condition is triggered, the digital twin is updated and all completed process tasks within a rolling freeze time window are locked, and the state change and the external event stream are updated in the current scheduling problem instance. The updated scheduling problem instance is transformed into an update objective function, and the first soft constraint weight, the second soft constraint weight, and the third soft constraint weight in the update objective function are updated as follows: ； ； In the formula, For the first The first, second, or third soft constraint weights after the second rescheduling iteration For the updated first soft constraint weight, second soft constraint weight, or third soft constraint weight, They represent the first sequence The current error of the second rescheduling iteration. These are the proportional action weight, integral action weight, and differential action weight, respectively. Update the weights of the operators: ； In the formula, For learning rate, For the weights of the updated operator, The weights of the operators before the update. This represents the objective function value of the operator this week. Based on the updated first soft constraint weights, second soft constraint weights, third soft constraint weights, and updated operator weights, an adaptive learning mechanism is used to solve the update objective function to obtain the updated scheduling plan and publish the updated scheduling plan.

8. A real-time scheduling system for PCB production based on hybrid intelligent decision-making, characterized in that, The system includes: The acquisition module is used to acquire PCB production information and construct a corresponding digital twin for the production information; The task module is used to identify explicit and implicit process tasks in the production work order based on the state of the digital twin. The constraint module is used to obtain the current business strategy and load hard constraint rule sets and soft constraint rule sets based on the current business strategy; The instance module is used to synthesize the explicit process tasks, the implicit process tasks, the hard constraint rule set, and the soft constraint rule set to obtain scheduling problem instances; The solution module is used to convert the scheduling problem instance into an objective function, perform multi-objective optimization on the objective function, and output an initial scheduling plan. The loop module is used to monitor the initial scheduling plan in real time and perform dynamic incremental rescheduling loops to complete the real-time scheduling of PCB production.

9. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the PCB production real-time scheduling method based on hybrid intelligent decision-making as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the real-time scheduling method for PCB production based on hybrid intelligent decision-making as described in any one of claims 1 to 7.

Images