A multi-line adaptive task cooperative scheduling method and system

Abstract

The application relates to the field of intelligent manufacturing, in particular to a multi-production-line self-adaptive task cooperative scheduling method and system. First, a multi-production-line self-adaptive task cooperative scheduling method is provided, which comprises the following steps: acquiring real-time production state data of each production line, and calculating a load pressure index of each production line according to the production state data; presetting a cooperative request threshold for each production line, receiving the load pressure index and the cooperative request threshold, and determining that a production line is an overloaded production line and extracting a task to be transferred when the load pressure index of the production line is not less than the cooperative request threshold; eliminating the overloaded production line to generate a candidate set, calculating a cooperative matching degree through a comprehensive benefit matching model of fused transfer cost, process compatibility and plan disturbance evaluation, and screening a task receiving production line; and generating an operation instruction to control the task transfer of the overloaded production line, the receiving production line and a material handling system. Secondly, a multi-production-line self-adaptive task cooperative scheduling system is provided according to the above content.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing of clothing, and in particular to a method and system for adaptive task collaborative scheduling across multiple production lines. Background Technology

[0002] With the evolution of the global manufacturing competition landscape and the diversification of consumer demands, flexible hanging garment production lines, as core equipment in modern garment manufacturing, enable production needs such as multi-variety, small-batch, and rapid delivery. In factory environments with multiple production lines operating in parallel, cross-production line task scheduling has become a key technical means to ensure production continuity in order to cope with sudden dynamic tasks (such as emergency order insertions) or unexpected interruptions (such as equipment failures or staff absences). However, how to achieve efficient and collaborative task transfer between different production lines under the constraints of multi-production line resources has become a major challenge facing current manufacturing systems.

[0003] Despite the existence of various production optimization solutions, existing technologies still have certain gaps in handling complex cross-production line collaborations. First, the collaborative triggering mechanism lacks scientific basis, relying heavily on fixed thresholds or human experience, making it difficult to achieve accurate early warning triggers based on data such as future working hours and real-time utilization rates. Second, the scheduling decision-making logic is one-sided; existing solutions struggle to balance the impact of local production line task injections on the overall plan, neglecting process compatibility limitations and logistical cost losses during transfers, resulting in cross-line scheduling decisions often failing to achieve global optimization.

[0004] Therefore, how to achieve intelligent, accurate, and systematic cross-production line collaborative scheduling remains a pressing technical problem that needs to be solved in the current informatization of the manufacturing industry. Summary of the Invention

[0005] This application first proposes a multi-production line adaptive task collaborative scheduling method to solve the above-mentioned problems. Secondly, this application proposes a multi-production line adaptive task collaborative scheduling system.

[0006] As a first aspect of this application, a method comprising the following steps is proposed: S1: Obtain real-time production status data for each garment production line, and calculate the load pressure index for each garment production line based on the production status data. S2: Set a preset collaboration request threshold for each garment production line, receive the load pressure index and the collaboration request threshold, and when the load pressure index of a certain garment production line is not less than the collaboration request threshold, determine that the garment production line is an overloaded production line and extract the tasks to be transferred from it. S3: Eliminate the overloaded production lines in all garment production lines and generate a candidate set. Calculate the coordination matching degree of each garment production line in the candidate set by integrating the comprehensive benefit matching model that evaluates transfer costs, process compatibility, and plan disturbance. Based on the coordination matching degree, select the matching task receiving production line from the candidate set. S4: Generate an operation instruction based on the relevant information of the task receiving production line and the task to be transferred. The operation instruction is used to control the overload production line, the task receiving production line and the material handling system to transfer the task to be transferred from the overload production line to the task receiving production line.

[0007] Optionally, the production status data includes the total remaining standard working hours of the queue, the real-time utilization rate of key equipment, the quantity of work-in-process, the rated daily production capacity, and the upper limit of work-in-process capacity.

[0008] Optionally, the formula for calculating the load pressure index (LPI) is as follows: in, For garment production lines, For time, The total remaining standard working hours in the queue. For rated daily production capacity, For the real-time utilization rate of key equipment, The quantity of work-in-process. This is the upper limit for work-in-process capacity. This is the weighting coefficient for queue workload. The weighting coefficient of real-time utilization rate of key equipment. The weighting coefficient for the work-in-process level is, and + + =1.

[0009] Optionally, step S2 can be further elaborated as follows: S21: Pre-set the response threshold for each garment production line. The response threshold includes a safety threshold, an early warning threshold, the collaboration request threshold, and an emergency circuit breaker threshold, and satisfies 0 < the safety threshold < the early warning threshold < the collaboration request threshold < the emergency circuit breaker threshold. S22: Receive the load pressure index of each garment production line and compare it with the response threshold; When the load pressure index of a certain garment production line is detected to be not less than the collaborative request threshold, the garment production line is determined to be an overloaded production line and tasks to be transferred are extracted from the overloaded production line. When the load pressure index of a certain garment production line is detected to be not less than the emergency circuit breaker threshold, an emergency warning signal is output and a preset emergency coordination plan is executed.

[0010] Optionally, step S22 further includes obtaining sample data of the historical load pressure index of the garment production line, wherein the initial value of the response threshold is set by the statistical quantile of the sample data; Wherein, the safety threshold is the 30th percentile of the sample data, the early warning threshold is the 70th percentile of the sample data, the collaborative request threshold is the 85th percentile of the sample data, and the emergency circuit breaker threshold is the 95th percentile of the sample data.

[0011] Optionally, in step S3, The candidate set is the set of production lines whose load pressure indices are all less than their respective safety thresholds after removing the overloaded production lines; when the candidate set is empty, the candidate set is the set of degraded production lines whose load pressure indices are all less than their respective warning thresholds after removing the overloaded production lines. The calculation of the comprehensive benefit matching model includes: Based on historical data, determine the minimum value of the transfer cost. and maximum value and the minimum planned disturbance time. and maximum value ; Estimated transfer cost for the candidate set and estimated planned disruption time Standardize the process to obtain standardized transfer costs. and standardized plan disturbance : ; ; in, For the candidate set; It is a positive constant; and The range of values ​​for all values ​​is [0,1]. The compatibility score is calculated using the following formula. : ; in, The score represents the process compatibility. The weighting coefficient for the transfer cost, The weighting coefficient for the process compatibility, Let be the planned disturbance coefficient, and + + =1.

[0012] Optionally, in step S3, the process compatibility score is obtained. To calculate the collaborative matching degree and the process compatibility score. The calculation method is as follows: In the pre-maintained process knowledge base, each task type is associated with a process requirement vector. ; Each production line in the pre-maintained production line resource file Each is associated with a resource capability vector. ; Define matching degree function According to each process requirement dimension In the middle, the process requirement vector With resource capability vector The degree of matching is divided into complete match, partial match, and no match, where complete match refers to... When partially matched When there is a mismatch ; The process compatibility score This is the weighted average of the matching degree of all key process requirements; When at least one necessary hard process requirement dimension matches the At that time, the process compatibility score is determined. .

[0013] Optionally, step S4 includes: S41: Generate a structured collaborative task package based on the task to be transferred. The collaborative task package has at least one key field, including a unique task identifier, a set of process documents, a bill of materials with storage location information, standard task hours, task priority, delivery deadline, source production line identifier, target production line identifier, and collaborative instruction generation timestamp. S42: Generate and issue multi-system linkage operation instructions based on the collaborative task package. The operation instructions set status checkpoints during operation. After the preceding instructions are confirmed to be successful, the operation instructions trigger subsequent instructions in sequence and provide timely feedback.

[0014] Optionally, step S5 may be included after step S4: S51: Receive the event reports generated from the collaborative events completed in steps S1-S4, and calculate the actual benefit value based on the event reports. The calculation formula is: ,in, This is the time cost equivalent conversion factor. This represents the total net order delay time avoided by the system as a whole. This refers to the actual material transfer cost; S52: Based on actual benefit value In addition to the triggering context, decision parameters, and execution results of collaborative events, a historical data warehouse for the event reports is constructed. S53: Based on the historical data warehouse, adjust the weight coefficients of the load pressure index, the weight coefficients of the comprehensive benefit matching model, and the collaborative request threshold through an optimization algorithm; S54: Monitor the weighting coefficients of the adjusted load pressure index, the weighting coefficients of the comprehensive benefit matching model, and the collaborative request threshold, and automatically roll back to the previous stable parameter version when a performance degradation is detected.

[0015] As a second aspect of this application, a multi-production-line adaptive task collaborative scheduling system is proposed for executing the aforementioned multi-production-line adaptive task collaborative scheduling method, and further includes, Data acquisition and calculation unit: used to acquire production status data of each garment production line in real time, and calculate the load pressure index of each garment production line based on the production status data; Collaborative triggering unit: used to set a collaborative request threshold for each garment production line; when the load pressure index of a certain garment production line is detected to be not less than the collaborative request threshold, the garment production line is determined to be an overloaded production line, and cross-line collaborative scheduling is triggered for the tasks to be transferred on the overloaded production line. Candidate screening and matching unit: used to screen candidate production lines whose load status meets preset conditions from all garment production lines; for each candidate production line, calculate its collaborative matching degree by integrating transfer cost, process compatibility and plan disturbance evaluation through a comprehensive benefit matching model; select the optimal task receiving production line from the candidate production lines based on the collaborative matching degree; Scheduling and execution unit: used to control the overload production line, the task receiving production line and the material handling system, and to perform the operation of transferring the task to be transferred from the overload production line to the task receiving production line for production.

[0016] The beneficial effects of this application are as follows: 1. The method proposed in this application calculates the load pressure index by acquiring production status data in real time, and automatically triggers task transfer when the garment production line is overloaded. Combined with the comprehensive benefit matching model, the optimal receiving production line is selected. This enables adaptive and collaborative scheduling of production tasks between different production lines, effectively balancing the load of each production line, avoiding production congestion or delays caused by local overload, and improving overall production efficiency and equipment utilization.

[0017] 2. By refining production status data into multiple dimensions such as the remaining standard working hours of the total queue, the real-time utilization rate of key equipment, and the quantity of work-in-process, and by using a specific load pressure index formula for quantitative calculation, it can comprehensively and accurately reflect the real load situation of the garment production line at different time periods, providing precise data support for triggering collaborative scheduling.

[0018] 3. By setting a response threshold system including safety threshold, early warning threshold, collaborative request threshold and emergency circuit breaker threshold, and initializing it based on the statistical quantile of historical data, the system can adapt to the characteristics and historical working conditions of different garment production lines, realize hierarchical early warning and response, ensure the timeliness of collaborative scheduling, avoid false triggering, and at the same time, the emergency circuit breaker mechanism provides protection for extreme situations, enhancing the robustness and adaptability of the method operation.

[0019] 4. When selecting production lines to receive tasks, by constructing a comprehensive benefit matching model that integrates transfer costs, process compatibility, and plan disturbance, and by using a specific quantitative calculation method for process compatibility, the target with the best comprehensive benefits can be selected from multiple candidate production lines, ensuring the feasibility of task transfer, while minimizing the additional costs and interference with the original production plan caused by the transfer.

[0020] 5. By using a pre-defined matching mechanism between process requirement vectors and resource capability vectors, the feasibility of cross-production line scheduling is ensured from a technical perspective. The introduction of a mandatory veto system for rigid process requirements effectively avoids scheduling failures or product quality incidents caused by unsupported production line functions, achieving a deep integration of production flexibility and process rigor, and guaranteeing the processing quality of cross-production line tasks.

[0021] 6. By generating structured collaborative task packages containing detailed information and setting status checkpoints and sequentially triggered operation instructions, efficient linkage between the overload production line, receiving production line, and material handling system is achieved, ensuring data consistency and operational reliability during task transfer and facilitating the tracking and management of the transfer process.

[0022] 7. By calculating the actual benefit value after task execution and building a historical data warehouse, the system continuously adjusts the weight coefficients of the load pressure index, the weight coefficients of the comprehensive benefit matching model, and the collaborative request threshold using optimization algorithms, and sets an automatic rollback mechanism, enabling the system to continuously learn and optimize itself, adapt to changes in the production environment, and maintain the efficiency and stability of the scheduling strategy in the long term.

[0023] 8. The system proposed in this application realizes the automated execution of the above-mentioned multi-production line adaptive task collaborative scheduling method through the collaborative work of each unit, and provides a full-process solution from data collection, trigger judgment, matching and filtering to scheduling execution, with the characteristics of intelligence, automation and high reliability. Attached Figure Description

[0024] Figure 1 This is a flowchart of the multi-production line adaptive task collaborative scheduling method in this specific embodiment. Detailed Implementation

[0025] The present application will be further described in detail below with reference to the accompanying drawings.

[0026] As the first aspect of this specific embodiment, such as Figure 1 As shown, a multi-production line adaptive task collaborative scheduling method is proposed, which includes the following steps: S1: Obtain real-time production status data for each garment production line and calculate the load pressure index for each garment production line based on the production status data. S2: Preset a collaboration request threshold for each garment production line, receive the load pressure index and the collaboration request threshold, and when the load pressure index of a certain garment production line is not less than the collaboration request threshold, determine that the garment production line is an overloaded production line and extract the tasks to be transferred from it. S3: Eliminate overloaded production lines in all garment production lines and generate a candidate set. Calculate the synergy matching degree of each garment production line in the candidate set by integrating transfer costs, process compatibility, and plan disturbance assessments using a comprehensive benefit matching model. Based on the synergy matching degree, select the matching task receiving production line from the candidate set. S4: Generate operation instructions based on the relevant information of the task receiving production line and the task to be transferred. The operation instructions are used to control the overload production line, the task receiving production line and the material handling system to transfer the task to be transferred from the overload production line to the task receiving production line.

[0027] The method proposed in this application calculates the load pressure index by acquiring production status data in real time, and automatically triggers task transfer when the garment production line is overloaded. Combined with the comprehensive benefit matching model, it selects the optimal receiving production line, which can realize adaptive collaborative scheduling of production tasks between different production lines, effectively balance the load of each production line, avoid production congestion or delays caused by local overload, and improve overall production efficiency and equipment utilization.

[0028] like Figure 1As shown, this scheduling method first collects real-time status data of the garment production line, integrates three dimensions—queue working hours, equipment utilization rate, and work-in-process level—to calculate the normalized load pressure index (LPI), and sets four-level response thresholds for each production line based on historical data statistical quantiles: safety, early warning, collaborative request, and emergency circuit breaker, thereby achieving refined hierarchical perception and triggering of overload risks. Secondly, when the load of a production line reaches the collaboration request threshold, the system will select a candidate set from all production lines with safe loads. By integrating three key factors—transfer cost, process compatibility, and plan disturbance—the system will quantitatively evaluate the collaboration matching degree of each candidate production line, thereby selecting the receiver with the best global benefit. Subsequently, the system will generate structured collaboration task packages and drive the linkage of multiple systems, such as the Manufacturing Execution System (MES) and Material Logistics System (WMS, AGV), in accordance with the principles of atomicity and sequence, to ensure the automated and reliable execution of task transfer. Finally, in some implementations, a data-driven adaptive optimization mechanism is introduced: a post-event quantitative benefit evaluation is performed after each collaboration, a historical data warehouse is built, and optimization algorithms (such as Bayesian optimization) are used to continuously adjust core parameters such as load weight, matching model weight, and threshold quantile, so that the scheduling strategy can continuously adapt to changes in the production environment and achieve long-term performance self-evolution. The following steps will be described in detail.

[0029] Step S1 specifically includes the following steps: S11: Real-time collection of production status data; The aforementioned production status data includes at least the total remaining standard working hours in the queue, the real-time utilization rate of key equipment, the quantity of work-in-process, the rated daily production capacity, and the upper limit of work-in-process capacity.

[0030] Specifically, the server uses a standardized application programming interface (API) to obtain production status data in real time from the manufacturing execution system (MES) or the field Internet of Things (IoT) monitoring terminal on the production line at preset intervals, such as every 10, 20, or 30 seconds.

[0031] Production status data includes: Total remaining standard working hours in the queue For each production line Collect the sum of the remaining standard working hours for all its unfinished production orders, and record it as... The unit is hours (h). The total remaining standard working hours in the queue is a core indicator used to measure the total amount of work to be completed in the future on the production line. Its data can be obtained from the order management and progress tracking module of the MES system.

[0032] Real-time utilization rate of key equipment Collection production line The real-time utilization rate of bottleneck processes or key equipment is denoted as... Bottleneck processes or key equipment can be identified through expert judgment, and this judgment information is then input into a server storing the multi-production-line adaptive task collaborative scheduling method of this application. The real-time utilization rate of key equipment is the ratio of actual equipment operating time to total statistical time, and its value range is... The real-time utilization data of key equipment comes from the equipment monitoring system or the equipment status module of the MES. When the value is close to 1, it indicates that the equipment is operating at near full capacity and has a small capacity buffer.

[0033] Work in progress Collection production line The quantity of all materials currently in the process of processing (work in progress) is denoted as follows. The unit is pieces. The work-in-process quantity data is calculated in real time by the MES system based on material input and work reporting records, and is used to reflect the material accumulation and flow smoothness within the production line.

[0034] Each production line The rated daily production capacity and the upper limit of work-in-process capacity can be obtained from the preset configuration database or basic files. The rated daily production capacity and the upper limit of work-in-process capacity are static capacity parameters.

[0035] in: Rated daily production capacity Production line The theoretical maximum daily output capacity under standard operating conditions is denoted as... The unit is standard working hours per day (h / day). Rated daily production capacity can be determined by the production line. The equipment is provided directly by the supplier or set based on industrial engineering (IE) analysis, theoretical cycle time of the equipment, or statistical averages of long-term historical production data to measure the baseline production capacity of the production line.

[0036] Work-in-process capacity limit Production line The production line was designed based on calculations of its physical layout length, intermediate workstations, or intermediate turnover capacity. The maximum allowable quantity of work-in-process is denoted as The unit is pieces. Maximum capacity of work-in-process inventory. It is a management threshold used to prevent production blockages and efficiency decline caused by excessive accumulation of work-in-process.

[0037] S12: Calculate the load pressure index (LPI) based on production status data.

[0038] Based on the dynamic data collected in step S11 and the obtained static parameters, for each production line time Calculate its load pressure index The load pressure index is calculated using a linear weighted formula, as follows: in: For garment production lines; For time; The total remaining standard working hours in the queue; Rated daily production capacity; Real-time utilization rate of key equipment; This refers to the quantity of work-in-process. This represents the upper limit of work-in-process capacity. , , These are weighting coefficients, corresponding to the contribution weights of the three dimensions—queue time load, real-time utilization rate of key equipment, and work-in-process level—to the overall LPI. Queue time load refers to the time consumed to complete all tasks currently in the queue. Its initial value can be set by domain expert experience combined with historical data. In some preferred embodiments, it can be: , , And satisfy the constraints. .at the same time, , , The specific weights can be dynamically optimized and adjusted by the system in step S5.

[0039] By , Divide by the reference parameter respectively and This enables the comparability and normalization of load indicators among production lines with different production capacities.

[0040] In one specific implementation, for production line P1, at a certain data acquisition time The system retrieved the following data: =80 hours =0.90 (i.e., 90%) =150 pieces. Its static parameters are known to be: =100 hours / day =200 items, using initial weights , , The calculation process for its load pressure index is as follows: =0.5*(80 / 100)+0.3*0.90+0.2*(150 / 200)=0.5*0.8+0.3*0.9+0.2*0.75=0.4+0.27+0.15=0.82 The calculation results show that the load pressure index of production line P1 at the current moment is 0.82.

[0041] By refining production status data into multiple dimensions such as the total remaining standard working hours in the queue, the real-time utilization rate of key equipment, and the quantity of work-in-process, and using a specific load pressure index formula for quantitative calculation, the actual load situation of the garment production line at different time periods can be comprehensively and accurately reflected, providing precise data support for triggering collaborative scheduling. Furthermore, due to the frequent and numerous order insertions on garment production lines, a load pressure index needs to be obtained within a short timeframe to quickly inform factory decisions on whether to adjust production allocation. Therefore, the load pressure index should not be set too complexly, otherwise it will reduce overall work efficiency.

[0042] In the implementation of the garment hanging production line, the cycle time of the hanging production line is relatively slow, with each station taking 3-8 minutes. The MES system's data acquisition cycle is 60 seconds per acquisition, and the acquisition interval itself has a smoothing effect, eliminating the need for more complex filtering algorithms (such as Kalman filtering, wavelet denoising, etc.). The load stress index (LPI) data for each factor comes from automatically acquired readings by the system, ensuring a stable data source without high-frequency noise interference. The system can perform a moving average filter (window = 5 acquisition cycles) on the data after acquisition to further eliminate instantaneous fluctuations.

[0043] In some implementations, step S2 specifically includes: S21: Pre-set the response threshold for each garment production line and set the initial value of the response threshold.

[0044] When the real-time load pressure index (LPI) of the garment production line reaches the response threshold, a tiered response process is triggered. In this specific embodiment, the response threshold is set to four levels, that is, for each production line... Four response thresholds are set, and the four response thresholds are: a security threshold, denoted as... The warning threshold is denoted as... The collaborative request threshold is denoted as... Emergency circuit breaker threshold, denoted as Meanwhile, the response threshold satisfies: 0 < .

[0045] Setting multi-level response thresholds can solve the problems of response lag or over-intervention that exist in traditional single fixed thresholds or manual experience-based judgment. By dividing the thresholds into four levels—safety, early warning, collaborative request, and emergency circuit breaker—the system can more finely classify the production line load status and execute differentiated response strategies. This setting avoids frequently triggering unnecessary collaborative scheduling to reduce costs during minor load fluctuations, while preserving a fast and automated response channel for genuine overload risks to ensure production, thus achieving precise control over scheduling timing. Furthermore, this configuration allows the system to adapt to the characteristics and historical operating conditions of different garment production lines, enabling tiered early warning and response. It ensures timely collaborative scheduling while avoiding false triggers, enhancing the robustness and adaptability of the method.

[0046] In some implementations, a statistical quantile method based on historical data is used for initialization and setting of initial values. These initial values ​​are set after the system is first deployed or after a major change in the production line configuration.

[0047] Specifically, the steps include the following: S211: The system collects the load pressure index of this production line at all historical moments over a representative period (e.g., the past 3 months). Sample data.

[0048] S212: Calculate the statistical quantiles of the sample data, and set the initial values ​​of each response threshold based on the quantile positions, where: The following explains each threshold in detail: Safety threshold : Set as history Sample Quantile values, in some implementations , This means that the load level was higher than this value for approximately 70% of the historical load pressure index of this production line. Therefore, when < At this time, the production line is under relatively light load and has sufficient capacity to accept external tasks.

[0049] Warning threshold : Set as history Sample Quantile values, in some implementations , This means that the load pressure index of the production line is higher than this value for about 30% of the time, indicating that there is a certain risk in the current state of the production line and an early warning is required.

[0050] Collaboration request threshold : Set as history Sample Quantile values, in some implementations , This means that the load level of the production line is higher than this value for about 15% of the time, indicating that the production line has a high risk of overload and external intervention is necessary.

[0051] Emergency circuit breaker threshold : Set as history Sample Quantile values, in some implementations , This means that the load level of the production line is higher than this value for about 5% of the time. At this time, the production line is in an extremely high load situation and requires unconventional measures (emergency coordination plan explained later) to deal with it quickly.

[0052] Among them, quantile parameters , , , System administrators or domain experts can preset settings based on the factory's risk appetite and production characteristics, and simultaneously adjust them. The value of will affect the evaluation of candidate receiving production lines. For example, adjusting it from 0.3 to 0.5 will relax the conditions for candidate receiving production lines, raising the safety threshold from the 30th percentile to the 50th percentile. This allows more production lines to be included in the candidate set, thus ensuring the availability of the collaborative scheduling mechanism even when the overall load is high. Additionally, adjusting... , , , It should still need to meet the requirements. In the preferred embodiment: , , , .

[0053] S22: Receive the load pressure index of each garment production line and compare it with the response threshold.

[0054] Real-time monitoring of each production line of It is then compared with a response threshold, and the corresponding tiered response is executed. Specifically: 1. When < When the production line is in a safe state, meaning its load is light, it is a qualified candidate to receive transfer tasks from other production lines. Production lines in this state can enter the candidate set for the subsequent step S3. 2. When ≤ < When the time is specified, it indicates that the production line is in a normal load state and is operating within a reasonable load range. Only routine monitoring is performed, and no scheduling actions are triggered.

[0055] 3. When ≤ < When the production line is in an early warning state, it needs to be marked as potentially overloaded. In a specific embodiment, a contingency plan calculation is initiated in the background, such as pre-screening possible collaborative receiving production lines and estimating the potential benefits of task transfer, to provide decision support for scheduling and buy decision time. In this state, a formal cross-line collaboration request does not need to be initiated immediately.

[0056] 4. When ≤ < When this occurs, it indicates that the production line is in a collaboration request state, and the production line is determined. The system has entered an overload state and requires cross-production line task coordination and scheduling. Once a production line is identified as overloaded, tasks to be transferred are extracted from it and then proceeded to step S3 for matching and decision-making.

[0057] 5. When ≥ When this occurs, it indicates that the production line is in an emergency shutdown state, meaning the production line is in an extremely urgent overload crisis. Step S3 needs to be skipped, and the following preset emergency coordination plan should be executed directly: Mandatory Recipient Designation: Based on a pre-maintained list of emergency standby production lines, assign tasks related to production lines in emergency shutdown states. Emergency standby production lines possess the following attributes: ① High process versatility (capable of handling multiple types of tasks); ② Reserved capacity or buffer time for rapid relocation; ③ Relatively central geographical location with controllable logistics costs. The list can be updated periodically based on production line status.

[0058] Simplify the process: Generate a minimal collaborative task package (containing only task identifier, bill of materials, and target production line) and send the highest priority enforcement instructions directly to the MES and logistics systems of the target production line.

[0059] Resource preemption: In some implementations, emergency coordination plans are authorized to suspend lower-priority tasks on the target production line to ensure the immediate insertion of emergency tasks. The priority is determined by the urgency priority rule in the following priority rules.

[0060] Post-event evaluation and calibration: All emergency circuit breaker operations will be recorded for post-event analysis to evaluate the effectiveness of the contingency plan and as a basis for adjustments. The basis for the threshold.

[0061] In a specific implementation described in step S12 above, assuming the 30%, 70%, 85%, and 95% quantiles of the historical LPI samples of production line P1 are 0.50, 0.75, 0.80, and 0.95 respectively, then its initial threshold is set as follows: =0.50, =0.75, =0.80, =0.95, if detected at a certain moment =0.82. Since it satisfies 0.80≤0.82<0.95, the system determines that P1 has entered the collaborative request state. P1 is an overloaded production line and extracts tasks to be transferred from it.

[0062] Compared with existing technologies that rely on a single fixed threshold or rely entirely on human experience, the above-described specific implementation method achieves the goal of minimizing unnecessary scheduling interventions while ensuring production safety, and retains a rapid processing channel for emergencies, significantly improving the scientific nature, robustness and efficiency of the scheduling system by finely classifying and differentiating the load status.

[0063] In some implementations, the detailed steps of step S22 above, which involves extracting tasks to be transferred from an overloaded production line, are as follows: when the production line is overloaded... After coordinated scheduling is triggered, one or more tasks are selected from the current production queue of the overloaded production line as tasks to be transferred. The following priority rules are used to select tasks to be transferred: 1. Prioritize based on urgency: Calculate the urgency of each task in the queue. Urgency coefficient: ,in, The delivery deadline timestamp for task m can be obtained directly from the ERP or MES system. The current system time. For safety buffer time. Select. The smallest (i.e., the most urgent) task is designated as the task to be transferred.

[0064] 2. Transfer Benefit Estimation: For multiple tasks with similar urgency, the potential comprehensive benefits of transferring them to each candidate production line can be estimated (refer to the S32 model below for rapid estimation), and the task with the most significant estimated comprehensive benefit improvement can be selected.

[0065] 3. Batch processing: If the overload is severe, the task with the longest standard working time or a group of tasks that can be processed in batches can be transferred to quickly reduce the production line load.

[0066] You can choose to enable one or a combination of the above strategies depending on the configuration. The finally selected task to be transferred is denoted as... Then proceed to the subsequent matching process.

[0067] In some implementations, step S3 specifically includes: When step S2 determines a certain production line After the production line is overloaded and collaborative scheduling is triggered, matching and decision-making are carried out: from the candidate set, the task receiver (task receiving production line) with the best comprehensive benefits is selected based on the comprehensive benefit matching model. The comprehensive benefit matching model quantitatively evaluates the three dimensions of economic cost, technical feasibility and plan stability involved in task transfer, so as to maximize the overall production benefits.

[0068] Specifically, the following steps are included: S31: Generate a candidate set.

[0069] Based on the real-time load status of each production line, garment production lines with the potential to receive tasks are selected to form a candidate set. Specifically: Assume the overload production line is Its pending transfer task is ; Traverse all non- Production line, referred to as production line For each production line Check its current load pressure index. Is it less than its own safety threshold? .

[0070] The conditions will be met. < All production lines Included in the candidate set, denoted as If the set at this time An empty value indicates that the entire plant is under high load, and the system automatically triggers a degradation search, relaxing the judgment criteria to... (Warning threshold) The candidate set is reorganized into a downgraded set where the load pressure index of the remaining candidate production lines is less than their respective warning thresholds after removing overloaded production lines.

[0071] In this step, only production lines with safe or light loads are considered as candidates for task reception (the degraded set). The degraded mechanism ensures that the scheduling algorithm remains available during peak seasons with high loads, thus avoiding global deadlock.

[0072] S32: For each candidate production line, calculate the synergy matching score using the comprehensive benefit matching model.

[0073] For the candidate set Each candidate production line Calculate the quantified collaborative matching score , Collaborative matching score The calculation is based on the estimated transfer cost, process compatibility score, and estimated planned disturbance time. At the same time, the estimated transfer cost and planned disturbance time are standardized between calculations to eliminate the influence of dimensions and make them comparable to the process compatibility score.

[0074] In this specific implementation, the calculation content of the comprehensive benefit matching model is as follows: The estimated transfer costs and planned disruption times are standardized using the following method: Determine the minimum transfer cost based on historical data or a pre-defined actual range. and maximum value and the minimum planned disturbance time. and maximum value ,but: in, It is a positive constant (usually very small, such as 0.001) used to prevent the denominator from being zero. and The value range is [0,1]. The larger the value, the better the performance of the item (lower cost and smaller disturbance).

[0075] Collaborative matching score The following formula is used to calculate: in, Indicates the estimated transfer cost: specifically, it represents... The required materials are from the overloaded production line. Transported to candidate production line The estimated additional economic costs incurred.

[0076] Estimated transfer costs It consists of two parts: the first part is the shortest logistics path distance between two production lines obtained based on the factory's digital twin model (a 3D model based on the actual size of the factory) or layout diagram, multiplied by the unit distance handling cost (which depends on logistics methods such as AGVs and manual carts); the second part is the standard preparation time cost required for task transfer (such as material inventory, re-hanging, etc.) calculated based on historical data statistics.

[0077] Coordinated scheduling itself incurs costs, and excessively high transfer costs may offset or even exceed the benefits of avoiding delays. Therefore, these costs need to be included in the evaluation model to maximize net benefits and avoid making unprofitable scheduling decisions. Meanwhile, the cost estimation model can be continuously calibrated in subsequent step S5 based on actual transfer cost data to improve its prediction accuracy.

[0078] Indicating process compatibility score: specifically representing Process requirements and candidate production lines The degree of matching between resource allocation (equipment, personnel skills, etc.).

[0079] In some implementations, process compatibility scores are obtained through a maintained process knowledge base and production line resource archives. Specifically, the process knowledge base refers to a vector of process requirements associated with each task type. Each element represents a key process requirement (e.g., required equipment model, personnel skill level, special auxiliary materials, processing precision, etc.), forming a production line resource file for each production line. Associate a resource capability vector This describes its capability value or Boolean value in this process dimension (e.g., list of available equipment, highest skill level, types of auxiliary materials available, etc.).

[0080] For each process requirement dimension Define the matching degree function : 1. Exact match: If equal to or belong to Within the scope of capabilities (e.g., the task requires "automatic cutting bed - model A", and the production line has this model of equipment), then .

[0081] 2. Partial matching: If and If alternative or downgraded solutions exist (e.g., different equipment models but similar functions, slightly lower skill levels but still sufficient), then (This value is configurable). The rules for determining partial matches need to be predefined in the process knowledge base (e.g., maintaining an "equipment substitution relationship table").

[0082] 3. Mismatch: If Completely not Support, then .

[0083] The process compatibility score is calculated using a veto mechanism. First, all hard process requirements marked as required are checked. If the match score of any hard requirement in the process requirement dimension is 0, then a process compatibility score is directly assigned. If all hard requirements are met, the production line will be removed from the available candidate pool; if all hard requirements are met, the process compatibility score will be calculated by weighted average of the matching degree of all key process requirements.

[0084] in, For the first The weight of each process requirement can be set by the process engineer based on the degree of impact of that requirement on task quality and efficiency. The default value is 1, and the score range is [value missing]. A higher value indicates better process compatibility. A mandatory process requirement dimension refers to a specific, uncompromising process requirement that must be met.

[0085] Ensuring that transferred tasks can be executed correctly and efficiently on the receiving production line through process compatibility scoring is a prerequisite for successful collaboration. Process compatibility score directly relates to production quality, efficiency, and potential rework risks. Incorporating process compatibility score into the model can systematically mitigate technical risks caused by equipment or skill mismatches. Its value range is... A higher score indicates better process compatibility. Furthermore, the introduction of a strict veto system for rigid process requirements effectively prevents scheduling failures or product quality incidents caused by unsupported production line functions, achieving a deep integration of production flexibility and process rigor, and ensuring the processing quality of cross-line tasks.

[0086] Indicates the estimated duration of the planned disruption: specifically, it will... Insert candidate production line The current production plan is used to estimate the total delay time caused by the original production plan. In this specific implementation, a simplified capacity-based model is used for estimation, such as: in, For production line The remaining working hours in the current queue. for Standard working hours For production line Rated daily production capacity This is the originally planned completion time.

[0087] pass Quantify the impact on the recipient's own production commitments (such as delivery dates for other orders) when providing assistance to other production lines. Smaller disruptions mean less negative impact on the overall factory delivery schedule. This factor is considered to achieve synergistic optimization of the global production plan, rather than localized firefighting.

[0088] , , These are weighting coefficients: corresponding to the relative importance weights of the three dimensions—transfer cost, process compatibility, and plan disturbance—in the final decision. By adjusting... , , Weighting can flexibly reflect the factory's management priorities at different times or under different order structures. For example, when more attention is paid to cost control, weighting can be increased. When focusing more on on-time delivery rate, it can be improved. .

[0089] It should be noted that, , , The range of values ​​is And satisfy the constraints. Preferably, the initial recommendation value is set based on experience and is [value missing]. =0.4, =0.4, =0.2. Prioritizing both economic feasibility (cost) and technical feasibility (process compatibility) simultaneously. =0.4, =0.4), and after ensuring that the task can be transferred and completed, optimize its perturbation to the global plan. The weight is slightly low (0.2).

[0090] The score represents the degree of collaboration matching: specifically, the candidate production line. Overall score For a dimensionless numerical value, a fraction The higher, the more likely it is to be Transfer to production line The better the expected overall benefits, the more likely the system will rank and select all candidate production lines based on this score.

[0091] S33: Select the optimal task receiving production line based on the coordination matching score.

[0092] After completing the review of all candidate production lines Collaborative matching score After calculation, the following decision logic is executed: 1. Sorting: Ranking the candidate set The production line in it is according to its Sort the fractions in descending order.

[0093] 2. Select: Will The candidate production line with the highest score is determined as the task receiving production line for this collaborative scheduling, and is denoted as production line. .

[0094] 3. Tie-breaking: If two or more candidate production lines have the highest... If the scores are the same, then further compare the current performance of these production lines. The production line with the lower LPI value (i.e., lighter real-time load) is selected as the [value]. This is to further ensure the production stability of the receiving party.

[0095] The decision module outputs the final selected task receiving production line. and their corresponding collaborative matching scores In addition, to enhance the transparency and explainability of decision-making, additional information can be attached. , , The specific values ​​of key evaluation factors provide dispatchers with clear decision-making basis and explanation.

[0096] In the above S12 example, a specific implementation is given, assuming the overloaded production line is P1, and the candidate set is... Tasks to be transferred The standard working hours are 10 hours. Based on model calculations: For P2: =50 yuan =0.9, =1.2 hours.

[0097] Known transfer cost range [ , =[30,100] yuan, disturbance time range [ , [0.5, 8] hours. (Take...) =0.001, calculate the standardized value: =(100-50) / (100-30+0.001)≈0.714; =(8-1.2) / (8-0.5+0.001)≈0.907. Take... =0.4, =0.4, =0.2, then: =0.4×0.714+0.4×0.9+0.2×0.907≈0.286+0.360+0.181=0.827.

[0098] For P5: =80 yuan =0.7, =3.0 hours. Therefore: =0.4×0.2857+0.4×0.7+0.2×0.6664=0.11428+0.28+0.13328≈0.528.

[0099] Compare =0.827 and =0.528, the system selects P2 as the task receiver for this collaboration.

[0100] As can be seen, compared with the decision-making based on only a single dimension (such as the current level of idleness) in the existing technology, by integrating the evaluation of three key dimensions—economic cost, technical feasibility, and plan stability—and the quantitative analysis with adjustable weights, it is possible to systematically avoid potential risks such as high-cost transfer, process mismatch, and serious plan disturbances. This allows each collaborative scheduling decision to pursue the maximization of global production benefits under multiple constraints, rather than local optimization.

[0101] When step S3 determines the optimal task receiving production line Subsequently, the system drives the relevant Manufacturing Execution System (MES) and Material Handling System (such as WMS and AGV) to work together to ensure the tasks to be transferred are completed. Able to safely, efficiently, and error-free transfer from overloaded production lines Migration to receiving production line And complete the synchronous update of the production plan.

[0102] Step S4 specifically includes the following steps: S41: Generate a structured collaborative task package based on the task to be transferred. The collaborative task package has at least one key field, including a unique task identifier, a set of process documents, a bill of materials with storage location information, standard task hours, task priority, delivery deadline, source production line identifier, target production line identifier, and collaborative instruction generation timestamp.

[0103] After triggering the collaboration, first encapsulate All relevant production information is used to generate standardized, machine-readable collaborative task packages. These collaborative task packages serve as the core data carrier for all subsequent execution instructions, employing a predefined structured data format (such as JSON or XML) to ensure seamless parsing between different systems.

[0104] Specifically, the annotations for the key fields included in the collaborative task package are as follows: Task unique identifier ( ): Originating from overloaded production lines The original work order number in the MES system is used to uniquely track the task throughout the entire collaborative process, ensuring data consistency.

[0105] Process document collection It can retrieve all the technical documents required for the task, such as cutting diagrams, sewing process sheets, and quality standard documents (usually in PDF or CAD format), to provide execution guidance for the receiving production line.

[0106] Bill of Materials with Storage Location Information List all materials required for the task (such as fabrics and accessories) and their current status on the overloaded production line. The specific bin location code of the in-process storage area. This information provides an accurate source address for material picking in the logistics system.

[0107] Task standard working hours ( ): Obtained from the process database and used to receive the production line for re-evaluating production capacity and scheduling production plans.

[0108] Task priority ( ): Can be inherited from the original plan or automatically assigned by the system according to the urgency of this collaboration (e.g., based on the value of the overloaded production line) and is used to guide the sorting decision of the receiving production line when inserting tasks.

[0109] Delivery deadline ( ): Obtained from the ERP or APS system and is a key constraint that the receiving production line scheduling needs to meet.

[0110] Source production line identifier ( ) and target production line identifier ( ): Respectively indicate the overloaded production line from which the task is removed and the receiving production line into which the task is moved and are used to guide the flow direction and path planning of all instructions.

[0111] Collaboration instruction generation timestamp ( ): Automatically generated by the system and is used for the timeliness analysis and problem tracing of the entire process.

[0112] It should be noted that regarding the key fields included in the collaboration task package, in some implementation manners, it includes all the above key fields, and in some other implementation manners, it can also be a combination of one or more of the above key fields.

[0113] Generating a structured collaboration task package can integrate the task information scattered in different systems into a complete and self-explanatory data entity, avoiding multiple and heterogeneous data queries and manual transmissions in the traditional mode, fundamentally reducing information errors and understanding ambiguities, providing a reliable and unified input source for the generation of subsequent automated instructions, and facilitating the tracking and management of the transfer process.

[0114] S42: Generate and issue operation instructions for multi-system linkage based on the collaboration task package. The operation instructions are configured to follow the principles of atomicity and sequentiality. Status checkpoints are set during the operation of the operation instructions. After the previous instruction is confirmed to be successful, the subsequent instructions are triggered in sequence and timely feedback responses are given to ensure the reliability of the entire process and the consistency of the data physical state.

[0115] In this specific implementation manner, the operation instructions are generated and issued in the following order: Instruction 1: Production suspension and plan removal.

[0116] Target system: Overloaded production line Manufacturing Execution System (MES).

[0117] Command content: ① Command MES to pause operation. ② Continue to feed or process related materials; It was officially removed from its current production queue, and the production line was recalculated. The remaining load and update .

[0118] Command 1 is triggered immediately after the collaborative task package is generated, serving as the starting point for all subsequent operations. After successful execution of Command 1, MES must return a confirmation signal that "task has been paused and plan has been updated".

[0119] Instruction 2: Execute material transfer.

[0120] Target systems: Material handling systems, such as warehouse management systems (WMS) and automated guided vehicle (AGV) scheduling systems.

[0121] Instruction content: According to Source locations in the (materials list) and (The target task receives information from the production line) and generates a specific material handling task order. The logistics system plans the optimal handling route and dispatches AGVs (Automated Guided Vehicles) or instructs personnel to move the specified materials from the production line. The temporary storage area was transported to the production line. The receiving area.

[0122] Instruction 2 is configured to rely on the successful confirmation of Instruction 1 to ensure that material handling only begins after material delivery has stopped, thus avoiding conflicts between data and physical status, such as the plan being deleted but the material still being processed.

[0123] After instruction 2 is successfully executed, the logistics system needs to return phased status feedback such as "materials have been dispatched", "in transit" and "delivered to the target storage location" until the material transfer is finally confirmed.

[0124] Instruction 3: Task reception and schedule insertion.

[0125] Target system: Receiving production line Manufacturing Execution System (MES).

[0126] Instruction content: ① Instruct the MES to receive and parse the collaborative task package; ② The MES, based on the task... , , and its current queue and production capacity The task is inserted into the optimal production position, and the insertion strategy should be as consistent as possible with the evaluation in step S3. Maintain consistency.

[0127] Instruction 3 is configured to rely on the "materials have been delivered" confirmation of Instruction 2 to ensure that the production plan is arranged only after the materials are in place, preventing situations such as the plan running in vain or there being no materials to cook. At the same time, after Instruction 3 is successfully executed, MES returns confirmation that the task has been received and the new production plan has been generated.

[0128] The key steps of pause-transfer-receive are configured as indivisible operation units, which either all succeed or can roll back to the state before execution in case of failure. At the same time, instruction 1, instruction 2 and instruction 3 are serial workflows, that is, instruction 2 depends on instruction 1 and instruction 3 depends on instruction 2, avoiding operation conflicts that may be caused by parallel operations. In addition, after the instruction is executed, it returns a clear and structured success or failure response, thereby realizing closed-loop control and enabling the system to perceive the execution results on site.

[0129] Furthermore, following step S42, step S43 is included: full-process status monitoring and closed-loop confirmation. In this specific embodiment, a collaborative dashboard is used to monitor and visualize the execution status of the instruction chain in real time, forming an execution closed loop.

[0130] Specifically, the dashboard displays the execution status of each instruction in real time and graphically (such as "pending issuance", "sent", "in execution", "success", "failure") as well as the location information of key materials, providing dispatchers with global visibility and centralized control points.

[0131] Meanwhile, each critical step has a preset timeout threshold (e.g., 5 seconds for MES command response, 30 minutes for AGV handling). If the command execution times out or returns a failure signal (e.g., AGV failure, MES order rejection), the system will automatically trigger an alarm and can perform a rollback (e.g., returning materials to the original storage location) or escalate to manual intervention according to preset strategies. By further setting timeout thresholds, the system can be prevented from waiting indefinitely due to a certain link being suspended, and problems can be detected in a timely manner. The timeout threshold can be set based on the statistical quantile of historical performance data (e.g., 95th percentile) to balance avoiding false alarms and timely problem detection.

[0132] In some implementations, the system is configured to determine that the collaborative scheduling execution was successful only after all three instructions (Instruction 1, Instruction 2, and Instruction 3) have received final success confirmations. Upon successful execution, a complete execution log (including timestamps of each stage, responsible person / equipment, and final status) can be immediately recorded, and the collaborative event can be officially closed. Simultaneously, the relevant central database can be updated to ensure data visibility across all systems. Figure 1 To (e.g., to The production line to which it belongs is marked as ).

[0133] Furthermore, after step S4, the system also includes steps to evaluate the execution effect and adaptively optimize the parameters, so as to quantitatively evaluate the actual effect of each collaborative scheduling, and use the accumulated historical data to automatically adjust the key decision parameters of the system (such as LPI weight, matching model weight, thresholds at all levels, etc.) through optimization algorithms, so that the system can continuously adapt to the dynamically changing production environment (such as changes in order structure and equipment efficiency fluctuations), continuously improve long-term scheduling efficiency, and achieve its self-growth and continuous optimization.

[0134] S5 specifically includes the following steps: S51: Receive the event reports generated from the collaborative events completed in steps S1-S4, and calculate the actual benefit value based on the event reports. The calculation formula is: ,in, This is the time cost equivalent conversion factor. This represents the total net order delay time avoided by the system as a whole. This refers to the actual material transfer cost; S52: Based on actual benefit value In addition, the triggering context, decision parameters, and execution results of collaborative events are used to build a historical data warehouse for event reporting; S53: Based on historical data warehouse, adjust the weight coefficients of load pressure index, comprehensive benefit matching model and collaborative request threshold through optimization algorithm; S54: Monitors the weighting coefficients of the adjusted load pressure index, the weighting coefficients of the comprehensive benefit matching model, and the collaborative request threshold. When a performance degradation is detected, it automatically rolls back to the previous stable parameter version.

[0135] In step S51, a post-hoc quantitative evaluation of the synergistic effect is performed when the receiving production line... Finish After production is completed and reported, the effectiveness evaluation process for this collaborative event is initiated to calculate the quantified actual benefit value. This serves as an indicator for evaluating the quality of this scheduling decision.

[0136] At the same time, predefine the time cost equivalent conversion factor. (Unit: Yuan / Hour) represents the equivalent economic value of avoiding one hour of order delay. This coefficient can be set by the finance department based on a combination of factors such as penalty clauses for order delays, loss of customer reputation, or opportunity costs, or it can be derived from historical data.

[0137] Actual benefit value The following formula is used to calculate: in, The total net avoided order delay time. This considers not only reducing the workload on overloaded production lines but also deducting the negative impact on receiving production lines. In this specific implementation, Can be triggered at the time of collaboration Simulations and estimations were performed after the task was completed: Pre-coordination delays: At any time, based on the production line The queue at that time Production capacity And the delivery dates for all orders, simulated calculations if not transferred. production line Total delay time for all orders .

[0138] Actual delays after coordination: After completion, based on the production line The actual production records are used to calculate the total actual delay time of the orders. .

[0139] but: The latter half represents the actual delay time caused by the insertion of tasks on the receiving production line. This demonstrates the positive benefits of collaboration, directly quantifying the contribution of collaborative scheduling in ensuring order delivery and improving plan stability. This directly corresponds to the most fundamental business objective of the scheduling system: reducing production delays. The value range is usually ≥0, and the unit is usually hours (h).

[0140] The actual material transfer cost represents the actual material transfer cost incurred in this collaboration. After step S4 is completed, the system extracts the actual consumption of this handling task from the operation records of the material logistics system (such as WMS warehouse management system or AGV automated guided vehicle system). This typically includes AGV operating costs, manual preparation time costs, and possible additional material handling costs.

[0141] This reflects the negative effects of collaboration, quantifying the actual economic costs incurred to achieve task transfer. Including these costs in the benefit calculation ensures a net benefit assessment, i.e., deducting execution costs, thus preventing suboptimal decisions that incur excessive costs to reduce minor delays from being misjudged as good decisions.

[0142] The value range is usually ≥0, and its unit is a monetary unit.

[0143] This represents the net benefit quantification value of a single collaborative event, serving as a core evaluation indicator and a learning signal for subsequent optimization. This indicates that the collaboration resulted in a net positive benefit (the value of the delays avoided was greater than the costs incurred). This indicates that the collaboration was counterproductive. The absolute value reflects the magnitude of the benefit or loss, through The abstract scheduling effect ("How well did this scheduling work?") can be transformed into a quantifiable and comparable single value, providing a clear and computable objective for subsequent parameter optimization: maximizing... .

[0144] In one specific implementation shown in step S12, assuming that during a single collaboration, the system estimates that if the task is not transferred, the overloaded production line P1 will generate... =50 hours of delay. After coordination, the actual delay of P1 is calculated as follows: =30 hours. The logistics system reports the actual transfer cost. =15 yuan, set the time cost equivalent conversion factor. =1 (unit: yuan / hour), substituting into the benefit formula: =1*20-15=5 yuan.

[0145] In step S52, a collaborative event historical data warehouse is constructed to store the triggering context, decision parameters, execution results, and actual benefit values ​​of each collaborative scheduling.

[0146] To support continuous learning, the system maintains a historical data warehouse for collaborative events, which structurally stores the complete context, decision data, and post-event evaluation results of each collaborative scheduling session. This data can be stored at fixed time intervals. Trigger a data collection and batch processing, for example Zhou or Months. Cycle. The choice of period needs to balance the timeliness of optimization with the statistical significance of data. If the period is too short, there will be few data samples and the optimization noise will be large; if the period is too long, the system will adapt to environmental changes slowly.

[0147] For the This collaborative event records, but is not limited to, the following data: Triggering context: Overloaded production line Triggering time The threshold values ​​at each level of the production line at that time ( , , , ).

[0148] Decision input: Candidate set and The estimated value used for matching degree calculation , , .

[0149] Decision parameters: Weighting parameters used in decision-making ( , , )and( , , ).

[0150] Decision result: Selected receiving production line k ∗ and its matching score .

[0151] Execution result: Actual transfer cost Delays to be avoided Calculated actual benefits .

[0152] Batch data aggregation: for a collection period Internal occurrence For each event, calculate the average actual benefit for that period: in, It is an indicator for measuring the overall scheduling performance of the system within a certain period.

[0153] By building a historical data warehouse, high-quality, structured foundational data is provided for parameter optimization. Empirical data is obtained by fully recording the decision input, decision output, and actual results of each collaborative process. (Average benefit over a period of time) rather than a single occurrence As a core performance indicator, it can prevent the system from adopting high-risk strategies in pursuit of high efficiency in a single instance.

[0154] In step S53, based on historical data in the historical data warehouse, the weight coefficients of the load pressure index, the weight coefficients of the comprehensive benefit matching model, and the collaborative request threshold are adjusted through an optimization algorithm.

[0155] At the end of each data collection period, the system automatically adjusts its core decision parameters using data from the current period and multiple historical periods, with the goal of maximizing the expected long-term average benefit. This can be formally expressed as: Find a set of parameters , so that the objective function Maximize the value while satisfying the constraints of each parameter (such as the sum of weights being 1), where The arithmetic mean of the actual benefits of all coordinated events within the period is the core KPI for measuring scheduling performance.

[0156] In some implementations, iterative optimization algorithms are used to solve the problem, which can be achieved through the following steps: 1. Construct the loss / reward function: Define the loss function associated with the objective function. (For gradient descent algorithms) or reward function (For reinforcement learning). The formula for the loss function related to the objective function is used because sample acquisition is costly and requires a long waiting period, and this formula supports a constrained parameter space.

[0157] 2. Parameter adjustment: Invoke the selected optimization algorithm and calculate parameters based on historical data. The direction and magnitude of the adjustment. The following algorithms can be selected based on the problem complexity and data volume: Gradient descent / Bayesian optimization: suitable for parameters with relatively small space. and In scenarios where the relationships are relatively smooth, the algorithm will attempt to fine-tune the parameters and observe historical data. The changing trend, and towards potential improvement The direction was adjusted.

[0158] Reinforcement learning (such as policy gradient): suitable for more complex dynamic environments. Each collaborative decision is viewed as an agent taking an action (selecting a receiving line) in a given state (the state of a garment production line). The actual benefits... As a reward, the optimal parameter strategy is obtained through continuous trial and error learning from historical data. .

[0159] The new parameter values ​​calculated by the optimization algorithm This information is updated in the system's runtime configuration for use in the next cycle's collaborative scheduling decisions.

[0160] During the optimization process, business logic constraints should be maintained (such as...) , and the relationship between threshold sizes At the same time, set reasonable safety boundaries for each parameter (such as...). ; This prevents the optimization process from generating parameters that are physically meaningless or that could lead to extreme system behavior.

[0161] In step S54, the optimization effect is monitored, and when performance degradation is detected, the system automatically rolls back to the previous stable parameter version.

[0162] To ensure the safety of the adaptive optimization process and prevent system performance degradation due to data noise or algorithmic issues, the following monitoring and protection mechanisms are established: A / B testing or shadow mode: When adding new parameters Before being formally applied to production scheduling, comparative testing (A / B testing) can be conducted on a small number of production lines or a specific time period, or the system can be run in shadow mode (i.e., parallel computation of decisions but without actual execution of instructions) to objectively evaluate its performance relative to the old parameters. Performance improvements.

[0163] Key Performance Indicator (KPI) Monitoring: Continuously monitor core business KPIs, such as periodic average efficiency. Parameters such as collaborative trigger frequency and average order delay time. If, after updating parameters, multiple consecutive periods... If a significant drop occurs or other abnormal fluctuations are triggered, the system will automatically trigger an alarm.

[0164] Automatic rollback mechanism: When the monitoring system detects severe performance degradation (such as...) A significant drop in performance will automatically trigger a rollback operation, restoring system parameters to the previous stable version. And immediately notify the administrator for manual inspection.

[0165] In some implementations, the value of the response threshold is configured to be dynamically adjusted according to the adaptive optimization mechanism of step S5, so that it can remain in line with the actual operating capacity of the production line, order structure and production environment.

[0166] Step S5 enables data-driven closed-loop optimization, allowing the system to proactively learn from historical records and automatically adjust parameters to the most suitable state for the current production environment. This continuously improves scheduling quality, helps the system cope with various emergencies, and ultimately achieves continuous evolution of intelligence and maximizes long-term production benefits.

[0167] The multi-production line adaptive task collaborative scheduling method of this application reduces production delays caused by untimely scheduling by more than 30% compared with traditional manual scheduling in practical applications. By avoiding unnecessary or high-cost transfers, it reduces ineffective scheduling overhead by about 15%.

[0168] By calculating the actual benefit value after task execution and building a historical data warehouse, the system continuously adjusts the weight coefficients of the load pressure index, the weight coefficients of the comprehensive benefit matching model, and the collaborative request threshold using optimization algorithms, and sets up an automatic rollback mechanism, enabling the system to continuously learn and optimize itself, adapt to changes in the production environment, and maintain the efficiency and stability of the scheduling strategy in the long term.

[0169] The scheduling method of this application will be explained in detail below based on an implementation method in a suit trousers production workshop: The suit trousers production workshop has eight flexible overhead production lines (P1-P8), all producing men's suit trousers, with a daily output of approximately 80-100 pairs and a production cycle of 3-8 minutes per workstation. All production lines have identical equipment configurations (five-thread overlock sewing machines, automatic pocket opening machines, etc.). However, due to differences in worker skill levels and the varying statistical distributions of historical load pressure indices, the safety thresholds and collaborative request thresholds calculated based on historical data differ between the production lines.

[0170] At 10:00 AM one morning, the system collected real-time production status data from MES, equipment PLCs, and hanging RFID tags for each production line, and calculated the real-time load pressure index LPI (the higher the value, the busier the production line): The security threshold and collaborative request threshold values ​​are calculated based on actual production statistics from the past three months.

[0171] P3's real-time LPI is 0.85, exceeding its collaboration request threshold of 0.75, thus indicating an overloaded production line. P3 currently has 6 pending orders in its queue, with order A having the longest standard working time (6 hours), and is therefore selected as the task to be transferred.

[0172] Candidate receiving production lines are selected based on their LPI (Local Perimeter) being lower than their own safety threshold before accepting new tasks: P2 (0.38<0.42), P4 (0.41<0.44), P6 (0.33<0.40), and P8 (0.36<0.43) meet the criteria. All four lines are in the same workshop and have the same equipment configuration.

[0173] Evaluate based on the following three practical factors: Transportation costs: The cost of manually moving trolleys from the P3 temporary storage area to each candidate production line is as follows: P2 is adjacent to P3, about 3 meters away, approximately 2 yuan; P4 is two workstations away, about 8 meters away, approximately 3 yuan; P6 is on the other side of the workshop, about 25 meters away, approximately 4 yuan; P8 is diagonally opposite P3, about 10 meters away, approximately 3 yuan. Due to the limited distance within the same workshop, the overall transportation costs are very low and the differences are not significant.

[0174] Process compatibility: All four candidate production lines have identical equipment configurations and are capable of completing Order A. The main difference lies in worker skill level; P2 and P6 have a higher proportion of skilled workers, while P4 and P8 have a slightly higher proportion of new workers, but all meet the process requirements for standard trouser production.

[0175] Planned Disruption: The impact of inserting order A on the original plans of each production line -- P2 currently has a small queue, and the latest order will be delayed by about 1.5 hours after insertion; P4 will be affected by about 0.8 hours; P6 already has a backlog, and will be affected by about 3 hours; P8 will be affected by about 1 hour.

[0176] Based on comprehensive assessment, P2 was selected – being adjacent to P3, it offered the most convenient handling, had skilled workers, and had manageable impact. The system generated a collaborative task package, and workers used carts to move the materials for order A from the P3 temporary storage area to P2 (approximately 2 minutes). The MES system on P2 automatically inserted the task into the production schedule. From the overload assessment to the completion of the transfer, approximately 10 minutes elapsed. Afterward, the backlog at P3 was reduced by about 15%, and order A was completed on time at P2.

[0177] In the aforementioned suit trousers production workshop, when an order is interrupted, it is necessary to quickly determine which line the order should be interrupted to ensure minimal impact on the overall production and maintain production schedules.

[0178] One afternoon at 2 PM, an urgent order was received – a VIP customer order for men's dress trousers, with a standard processing time of 20 hours (for the entire order, including multiple identical pieces), to be shipped the next day. The system assigned this order to P6, which currently had the lightest workload (real-time LPI=0.33, with 3 pending orders in the queue, totaling 14.8 hours of processing time).

[0179] Given the rated daily capacity of the P6 production line =38, Maximum capacity of work-in-process inventory =50 items. After the order is added to the P6 queue, the system collects the real-time utilization rate of the key equipment in P6 at the current moment. =0.51, Work-in-process quantity =22 pieces, the total remaining standard working hours Q(t) in the queue increased from 14.8 hours to 34.8 hours. After the order was added to queue P6, the total remaining standard working hours in the queue... The time increased from 14.8 hours to 34.8 hours. Substituting this into the LPI calculation formula: LPI_P6 = 0.5×(34.8 / 38) + 0.3×0.51 + 0.2×(22 / 50) = 0.458 + 0.153 +0.088 =0.699 The collaboration request threshold for P6 is 0.68. 0.699 ≥ 0.68 triggers overload.

[0180] The system determines that P6 is an overloaded production line and extracts tasks to be transferred from the P6 queue—high-urgency orders (VIP customers)—and prioritizes their production on P6. Following step S22 above, urgency is selected as the priority rule. Among the existing orders on P6, order B has a standard working time of 3 hours and a remaining delivery time of 28 hours. =28), the lowest urgency level, selected as the task to be transferred.

[0181] Candidate receiving production lines were selected: P2 (0.38<0.42), P4 (0.41<0.44), and P8 (0.36<0.43) met the requirements. P2 is adjacent to P6 (approximately 6 meters, transportation cost approximately 2 yuan), has complete equipment, skilled workers, and an acceptable delay of approximately 0.8 hours after insertion. Therefore, it was selected as the receiving production line.

[0182] Workers push carts to move materials for Order B from P6 to P2 (approximately 3 minutes), and the P2 system automatically schedules production. From the time the order is inserted to the completion of the transfer, it takes approximately 10 minutes. The inserted order is produced normally in P6 and completed the following morning; Order B is completed normally in P2, and the latest order originally scheduled for P2 is delayed by approximately 0.8 hours.

[0183] The production line parameters in the two embodiments above are based on relevant data from actual flexible garment hanging production lines. The safety threshold (30th percentile) and collaboration request threshold (85th percentile) for each production line are derived from their respective historical LPI data over the past 3 months. Different production lines have different thresholds due to different historical load characteristics. The weights of each LPI factor (queue hours 0.5, equipment utilization 0.3, work-in-process 0.2) cover the three dimensions of production line load: "future workload - current urgency - internal smoothness". The values ​​of α, β, and γ are initial empirical values ​​and can be dynamically adjusted through S5's adaptive optimization.

[0184] In the two implementation scenarios described above, the equipment configuration is identical within the same workshop and product line. Therefore, the differences in process compatibility are not significant. However, in actual production, if collaborative scheduling between different product lines is involved, process compatibility will become an important consideration.

[0185] As a second aspect of this specific embodiment, a multi-production-line adaptive task collaborative scheduling system is proposed for executing the above-described multi-production-line adaptive task collaborative scheduling method, including: Data acquisition and calculation unit: used to acquire production status data of each garment production line in real time, and calculate the load pressure index of each garment production line based on the production status data; Collaborative Trigger Unit: Used to set a collaborative request threshold for each garment production line; when the load pressure index of a certain garment production line is detected to be not less than the collaborative request threshold, the garment production line is determined to be an overloaded production line, and cross-line collaborative scheduling is triggered for the tasks to be transferred on the overloaded production line. Candidate screening and matching unit: used to screen candidate production lines whose load status meets preset conditions from all garment production lines; for each candidate production line, its collaborative matching degree is calculated by a comprehensive benefit matching model that integrates transfer cost, process compatibility and plan disturbance evaluation; and the optimal task receiving production line is selected from the candidate production lines based on the collaborative matching degree. Scheduling and execution unit: Used to control the overload production line, task receiving production line and material handling system, and to perform the operation of transferring the task to be transferred from the overload production line to the task receiving production line for production.

[0186] When the above-mentioned multi-production line adaptive task collaborative scheduling method includes step S5, the multi-production line adaptive task collaborative scheduling system also includes an adaptive optimization unit. The adaptive optimization unit is used to calculate the actual benefit value based on the execution result of the collaborative scheduling, build a historical data warehouse, and adjust the weight coefficient of the load pressure index, the weight coefficient of the comprehensive benefit matching model, and the collaborative request threshold through the optimization algorithm.

[0187] The system proposed in this application achieves automated execution of the above-mentioned multi-production line adaptive task collaborative scheduling method through the collaborative work of each unit. It provides a full-process solution from data acquisition, trigger judgment, matching and filtering to scheduling execution, and has the characteristics of intelligence, automation and high reliability.

[0188] This application discloses a multi-production-line adaptive task collaborative scheduling system and its various components, which can be implemented through software programs, hardware, or a combination of both. In this embodiment, each component is named and described in terms of its functional architecture. It should be understood that in practical applications, these components can be software systems carried on computer-readable media, or functional modules integrated into servers, terminal devices, or embedded processors. This functional description aims to cover all computing system architectures capable of executing the corresponding algorithm flow, and should not be limited to specific software or hardware implementations.

[0189] The above are merely embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made using the content of the present invention specification, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A multi-production line adaptive task collaborative scheduling method, characterized in that: Includes the following steps: S1: Obtain real-time production status data for each garment production line, and calculate the load pressure index for each garment production line based on the production status data. S2: Preset a collaboration request threshold for each garment production line, receive the load pressure index and the collaboration request threshold, and when the load pressure index of a certain garment production line is not less than the collaboration request threshold, determine that the garment production line is an overloaded production line and extract the tasks to be transferred from it. S3: Eliminate the overloaded production lines in all garment production lines and generate a candidate set. Calculate the coordination matching degree of each garment production line in the candidate set by integrating the comprehensive benefit matching model that evaluates transfer costs, process compatibility, and plan disturbance. Based on the coordination matching degree, select the matching task receiving production line from the candidate set. S4: Generate an operation instruction based on the relevant information of the task receiving production line and the task to be transferred. The operation instruction is used to control the overload production line, the task receiving production line and the material handling system to transfer the task to be transferred from the overload production line to the task receiving production line. The production status data includes the total remaining standard working hours of the queue, the real-time utilization rate of key equipment, the quantity of work-in-process, the rated daily production capacity, and the upper limit of work-in-process capacity. The formula for calculating the load pressure index (LPI) is as follows: in, For garment production lines; For time; The total remaining standard working hours in the queue; Rated daily production capacity; Real-time utilization rate of key equipment; This refers to the quantity of work-in-process. This represents the upper limit of work-in-process capacity. This is the weighting coefficient for queue workload. The weighting coefficient of real-time utilization rate of key equipment. The weighting coefficient for the work-in-process level is, and + + =1; The specific details of step S2 are as follows: S21: Pre-set the response threshold for each garment production line. The response threshold includes a safety threshold, an early warning threshold, the collaboration request threshold, and an emergency circuit breaker threshold, and satisfies 0 < the safety threshold < the early warning threshold < the collaboration request threshold < the emergency circuit breaker threshold. S22: Receive the load pressure index of each garment production line and compare it with the response threshold; When the load pressure index of a certain garment production line is detected to be not less than the collaborative request threshold, the garment production line is determined to be an overloaded production line and tasks to be transferred are extracted from the overloaded production line. When the load pressure index of a certain garment production line is detected to be not less than the emergency circuit breaker threshold, an emergency warning signal is output and a preset emergency coordination is executed. The calculation of the comprehensive benefit matching model includes: Based on historical data, determine the minimum value of the transfer cost. and maximum value and the minimum planned disturbance time. and maximum value ; Estimated transfer cost for the candidate set and estimated planned disruption time Standardize the process to obtain standardized transfer costs. and standardized plan disturbance : ； ； in, For the candidate set; It is a positive constant; and The value range is [0,1]; the candidate set is the set of production lines whose load pressure index is less than their respective safety thresholds after removing the overloaded production line; when the candidate set is empty, the candidate set is the downgrade set of production lines whose load pressure index is less than their respective warning thresholds after removing the overloaded production line. The compatibility score is calculated using the following formula. : ； in, The score represents the process compatibility. The weighting coefficient for the transfer cost, The weighting coefficient for the process compatibility, Let be the planned disturbance coefficient, and + + =1.

2. The scheduling method as described in claim 1, characterized in that, Step S22 further includes obtaining sample data of the historical load pressure index of the garment production line, and the initial value of the response threshold is set by the statistical quantile of the sample data. Wherein, the safety threshold is the 30th percentile of the sample data, the early warning threshold is the 70th percentile of the sample data, the collaborative request threshold is the 85th percentile of the sample data, and the emergency circuit breaker threshold is the 95th percentile of the sample data.

3. The scheduling method as described in claim 1, characterized in that: In step S3, the process compatibility score is obtained. To calculate the collaborative matching degree and the process compatibility score. The calculation method is as follows: In the pre-maintained process knowledge base, each task type is associated with a process requirement vector. ; Each production line in the pre-maintained production line resource file Each is associated with a resource capability vector. ; Define matching degree function According to each process requirement dimension In the middle, the process requirement vector With resource capability vector The degree of matching is divided into complete match, partial match, and no match, where complete match refers to... When partially matched When there is a mismatch ; The process compatibility score This is the weighted average of the matching degree of all key process requirements; When at least one necessary hard process requirement dimension matches the At that time, the process compatibility score is determined. .

4. The scheduling method as described in claim 1, characterized in that: Step S4 includes: S41: Generate a structured collaborative task package based on the task to be transferred. The collaborative task package includes at least one key field, which includes a unique task identifier, a set of process documents, a bill of materials with storage location information, standard task hours, task priority, delivery deadline, source production line identifier, target production line identifier, and collaborative instruction generation timestamp. S42: Generate and issue multi-system linkage operation instructions based on the collaborative task package. The operation instructions set status checkpoints during operation. After the preceding instructions are confirmed to be successful, the operation instructions trigger subsequent instructions in sequence and provide timely feedback.

5. The scheduling method as described in claim 1, characterized in that, Step S5 is included after step S4: S51: Receive the event reports generated from the collaborative events completed in steps S1-S4, and calculate the actual benefit value based on the event reports. The calculation formula is: ,in, This is the time cost equivalent conversion factor. This represents the total net order delay time avoided by the system as a whole. This refers to the actual material transfer cost; S52: Based on actual benefit value In addition to the triggering context, decision parameters, and execution results of collaborative events, a historical data warehouse for the event reports is constructed. S53: Based on the historical data warehouse, adjust the weight coefficients of the load pressure index, the weight coefficients of the comprehensive benefit matching model, and the collaborative request threshold through an optimization algorithm; S54: Monitor the weighting coefficients of the adjusted load pressure index, the weighting coefficients of the comprehensive benefit matching model, and the collaborative request threshold, and automatically roll back to the previous stable parameter version when a performance degradation is detected.

6. A multi-production-line adaptive task collaborative scheduling system, characterized in that: Used to execute the scheduling method as described in any one of claims 1 to 5 include, Data acquisition and calculation unit: used to acquire production status data of each garment production line in real time, and calculate the load pressure index of each garment production line based on the production status data; Collaborative triggering unit: used to set a collaborative request threshold for each garment production line; when the load pressure index of a certain garment production line is detected to be not less than the collaborative request threshold, the garment production line is determined to be an overloaded production line, and cross-line collaborative scheduling is triggered for the tasks to be transferred on the overloaded production line. Candidate screening and matching unit: used to screen candidate production lines whose load status meets preset conditions from all garment production lines; for each candidate production line, its synergistic matching degree is calculated by a comprehensive benefit matching model that integrates transfer costs, process compatibility and planning disturbance evaluation. Based on the aforementioned collaborative matching degree, the optimal task receiving production line is selected from the candidate production lines; Scheduling and execution unit: used to control the overload production line, the task receiving production line and the material handling system, and to perform the operation of transferring the task to be transferred from the overload production line to the task receiving production line for production.

Images