A production scheduling method, apparatus, equipment and medium based on digital twins

Abstract

This invention discloses a production scheduling method, apparatus, equipment, and medium based on digital twins, relating to the field of intelligent manufacturing. The method includes: collecting production data, generating an initial production state, defining initial production scheduling constraints, and constructing a digital twin model; predicting available production capacity using the digital twin model, identifying risk periods, and updating the twin state of production resources; coupling the predicted available production capacity with an order set to generate a task priority queue and update the production scheduling constraints; performing scheduling optimization and simulating conflict detection within the digital twin model; if conflicts exist, performing local rearrangement based on the conflict information until requirements are met, generating an executable scheduling scheme; distributing the executable scheduling scheme to production and logistics links, monitoring execution deviations, rescheduling affected tasks, completing production tasks, and updating the digital twin database. This invention improves the real-time performance and accuracy of scheduling responses.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing, and in particular to a production scheduling method, apparatus, equipment and medium based on digital twins. Background Technology

[0002] In the manufacturing process, production scheduling is a crucial link in achieving optimal resource allocation and on-time order delivery. Traditional production scheduling methods are mostly based on rule-driven, static models, or historical experience for scheduling decisions, using the production scheduling module in the manufacturing execution system to achieve initial task allocation and scheduling control. These methods typically employ fixed parameters and constraints, relying on heuristics, genetic algorithms, and constraint satisfaction to generate scheduling schemes based on known resource status and order requirements. The scheduling and execution loop is completed with the support of production management systems, equipment control systems, and information acquisition systems. Furthermore, recent research has attempted to introduce real-time sensing data into static models, using feedback control to adjust responses to abnormal states, thereby improving the robustness and effectiveness of the scheduling system.

[0003] However, in a dynamic manufacturing environment, traditional production scheduling methods still face certain challenges in terms of dynamic adaptability and multi-source data fusion. On the one hand, the quantitative prediction of future short-term available capacity and the explicit solidification of risk periods are relatively limited, making it difficult to proactively avoid bottlenecks before production scheduling. On the other hand, constraints and priorities are mostly statically set, making it difficult to write back conflict information and drive local rearrangement and rescheduling in the same round, resulting in strategy updates lagging behind the evolution of the physical process. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a production scheduling method based on digital twins to solve the problems of insufficient forward-looking identification of capacity fluctuations and insufficient dynamic adaptation of scheduling constraints to real-time states in the prior art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a production scheduling method based on digital twins, comprising,

[0008] Collect production data, generate initial production status, define initial production scheduling constraints, and build a digital twin model;

[0009] Predict the available capacity of production resources through digital twin models, identify risk periods, and update the twin status of production resources.

[0010] Couple the predicted available production capacity with the order set issued by the production management link, calculate the overall priority of the orders and generate a task priority queue, solidify the risk window and update the initial production scheduling constraints;

[0011] Based on the task priority queue and the updated production scheduling constraints, scheduling optimization is performed in the digital twin model to generate candidate scheduling schemes. The candidate scheduling schemes are then subjected to time-series simulation to detect conflicts in the digital twin environment.

[0012] Based on the detected conflicts, write back the conflict information to the updated production scheduling constraints, perform local rearrangement within the same round until the conflicts are eliminated and the performance indicators meet the preset requirements, and output an executable scheduling scheme.

[0013] The executable scheduling plan is distributed to the production and logistics links, execution deviations are monitored, local work condition packages are generated, affected tasks are rescheduled, the overall production task is completed, and production data is written back to the digital twin database to update the production resource available capacity forecast data, production scheduling constraints and optimization weights.

[0014] As a preferred embodiment of the production scheduling method based on digital twins described in this invention, wherein:

[0015] The construction of the digital twin model refers to collecting production data and aligning it with timestamps to generate a production dataset, generating an initial production state based on the production dataset, defining initial production scheduling constraints, and constructing a digital twin model by combining virtual production line topology, resource nodes, processing nodes, and logistics paths.

[0016] As a preferred embodiment of the production scheduling method based on digital twins described in this invention, wherein:

[0017] The specific steps for predicting the available production capacity of production resources, identifying risk periods, and updating the twin status of production resources are as follows:

[0018] Based on the digital twin model, the available capacity of production resources is predicted, and the prediction results of the available capacity of production resources are generated.

[0019] Based on the predicted available production capacity of generated resources, time periods below the production capacity threshold are identified as risk periods, and the twin status of production resources is updated.

[0020] As a preferred embodiment of the production scheduling method based on digital twins described in this invention, the steps of coupling the predicted available production capacity with the order set issued by the production management process, calculating the comprehensive priority of the orders and generating a task priority queue, fixing the risk window and updating the initial production scheduling constraints are as follows:

[0021] The production resource availability forecast results are coupled with the order set issued by the production management process to generate the production resource and order set coupling result;

[0022] Calculate the overall order priority for each order, and generate the overall order priority for all orders;

[0023] Sort orders by overall priority from highest to lowest, and generate a task priority queue;

[0024] Solidify the risk window and update the initial production scheduling constraints.

[0025] As a preferred embodiment of the production scheduling method based on digital twins described in this invention, wherein:

[0026] The execution scheduling optimization involves generating candidate scheduling schemes and performing time-series simulation to detect conflicts. The specific steps are as follows:

[0027] Perform scheduling optimization operations in the digital twin model to generate a set of scheduling optimization operations;

[0028] In a digital twin environment, a time-series simulation of the scheduling optimization computation set is performed to generate simulation results for candidate scheduling schemes;

[0029] The simulation results of candidate scheduling schemes are used to detect processing sequence conflicts, resource competition conflicts, and logistics blockage conflicts, and conflict detection results are generated.

[0030] As a preferred embodiment of the production scheduling method based on digital twins described in this invention, wherein:

[0031] The process involves writing back conflict information to the updated production scheduling constraints, performing local rearrangement until the conflicts are resolved and performance indicators meet preset requirements, and then outputting an executable scheduling scheme. The specific steps are as follows:

[0032] Based on the conflict detection results, the conflict type, conflict task identifier, conflict resource identifier, as well as the occurrence time and location are extracted and the conflict information is generated.

[0033] Write back the conflict information to the updated production scheduling constraints, and generate the updated production scheduling constraints result including the conflict constraints;

[0034] Based on the updated production scheduling constraint results based on conflict constraints, a local rearrangement operation is performed within the same round to generate the local rearrangement operation result.

[0035] Based on the results of the local rearrangement operation, the timing simulation is executed again to repeatedly check whether the conflict is eliminated and whether the performance indicators meet the preset requirements. When the conditions are met, an executable scheduling scheme is generated.

[0036] As a preferred embodiment of the production scheduling method based on digital twins described in this invention, wherein:

[0037] The process involves distributing the executable scheduling plan to the production and logistics stages, monitoring execution deviations, generating local work condition packages, rescheduling affected tasks, completing the overall production task, writing production data back to the digital twin database, and updating production resource availability forecast data, production scheduling constraints, and optimization weights. The specific steps are as follows:

[0038] The executable scheduling plan is distributed to the production and logistics stages, the execution deviations during the execution of the executable scheduling plan are monitored, the execution deviation results are generated, and local work condition packages are generated based on the execution deviation results.

[0039] Perform a rescheduling operation and generate the rescheduling result;

[0040] Complete the overall production task and generate the production task completion result;

[0041] Production data is written back to the digital twin database to update the production resource availability forecast data, production scheduling constraints, and optimization weights, thereby generating an updated digital twin database.

[0042] Secondly, the present invention provides a production scheduling device based on digital twins, comprising,

[0043] The data twin building module is used to collect production data, generate the initial production state, define the initial production scheduling constraints, and build a digital twin model.

[0044] The capacity forecasting and risk control module is used to forecast the available capacity of production resources, identify risky periods, and update the twin status of production resources.

[0045] The order priority queue module is used to couple the predicted available production capacity with the order set issued by the production management link, calculate the overall priority of the orders and generate the task priority queue, solidify the risk window and update the initial production scheduling constraints.

[0046] The simulation conflict detection module is optimized. Based on the task priority queue and the updated production scheduling constraints, scheduling optimization is performed in the digital twin model to generate candidate scheduling schemes. The candidate scheduling schemes are then subjected to time-series simulation to detect conflicts in the digital twin environment.

[0047] The conflict writeback and reordering module is used to write back conflict information to the updated production scheduling constraints, perform local reordering in the same round until the conflict is eliminated and the performance indicators meet the preset requirements, and output an executable scheduling scheme.

[0048] The execution monitoring and rescheduling module is used to distribute executable scheduling plans to the production and logistics links, monitor execution deviations, generate local work condition packages, reschedule affected tasks, complete the overall production tasks, write production data back to the digital twin database, and update production resource availability capacity forecast data, production scheduling constraints, and optimization weights.

[0049] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the production scheduling method based on digital twins as described in the first aspect of the present invention.

[0050] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the production scheduling method based on digital twins as described in the first aspect of the present invention.

[0051] The beneficial effects of this invention are as follows: By predicting the available capacity of production resources, identifying risk periods, and updating the twin state of production resources based on a digital twin model, it achieves forward-looking perception of future capacity fluctuations before production scheduling, dynamically avoids potential risks during scheduling, and ensures the continuity and reliability of resource allocation; by coupling the predicted available capacity of production resources with the order set issued by the production management link, calculating the comprehensive priority of orders and generating a task priority queue, solidifying the risk window and updating the initial production scheduling constraints, it achieves deep matching and ranking of order demand and resource capacity, carries out scheduling optimization and local reordering in the same round in the digital twin model, and outputs an executable solution through an iterative closed loop of "simulation-conflict detection-constraint write-back", improving the accuracy and timeliness of scheduling. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a flowchart of a production scheduling method based on digital twins.

[0054] Figure 2 This is a schematic diagram of a production scheduling system based on digital twins.

[0055] Figure 3 A flowchart for digital twin prediction and task priority queue generation.

[0056] Figure 4The flowchart for scheduling optimization—simulation—conflict resolution and rescheduling is shown. Detailed Implementation

[0057] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0058] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0059] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0060] Example 1, referring to Figures 1-4 As one embodiment of the present invention, this embodiment provides a production scheduling method based on digital twins, including the following steps:

[0061] S1. Collect production data, generate the initial production state, define the initial production scheduling constraints, and build a digital twin model.

[0062] Collect production data and generate production datasets.

[0063] Furthermore, by integrating and calling the equipment operation interface, distributed sensor acquisition devices, manufacturing execution management information sources, and logistics status acquisition devices, production data is collected comprehensively.

[0064] Production data includes operating parameters of production equipment, work-in-process status of materials, inventory levels, status of logistics nodes, and energy consumption data.

[0065] Furthermore, the collected production data is aligned according to timestamps to generate a production dataset.

[0066] The initial production state is generated based on the production dataset.

[0067] The initial production status includes the resource availability matrix, process completion rate, inventory vector, and logistics in transit.

[0068] Furthermore, based on the production dataset, the resource availability matrix, process completion ratio vector, inventory vector, and logistics in-transit quantity vector are calculated.

[0069] Resource availability is represented as:

[0070] ;

[0071] in, Indicates the first Class resources in the The availability of each process step Indicates the first Class resources in the The available time for each process step Indicates the first Class resources in the The time required for each process step.

[0072] Determine the rows and columns (resources × processes) and statistical period of the resource availability matrix. For each "resources × processes" combination, obtain the planned required time and available time. Calculate and fill the matrix cell by dividing the available time by the planned required time, and make corrections for 0, missing, out-of-bounds, and risk window situations to generate the resource availability matrix.

[0073] It should be noted that the process completion ratio vector is calculated and obtained through real-time task execution status data collected from the equipment operation interface and the manufacturing execution management information source.

[0074] Inventory vectors are obtained from material inbound and outbound records in the Manufacturing Execution Management information source and from inventory data in the warehousing system.

[0075] The logistics in-transit quantity vector obtains the quantity of goods that have not yet been transported by comparing the material loading departure time with the estimated arrival time.

[0076] Furthermore, based on the resource availability matrix, combined with the process completion ratio vector, inventory vector, and logistics in-transit quantity vector, the initial production state is generated.

[0077] Using the virtual production line topology as the framework, the resource availability matrix, process completion ratio vector, inventory vector, and logistics in-transit quantity vector are aligned by node / process, and the time window and unit are unified to generate the physical production line state vector at t=0.

[0078] Furthermore, based on the initial production state, initial production scheduling constraints are defined.

[0079] It should be noted that the initial production scheduling constraints include resource capacity limits, process routes, equipment maintenance plans, raw material supply delays, and energy ceilings.

[0080] Based on the rated capacity of the equipment, the upper limit of resource capacity is formed as a capacity constraint by reducing the equipment maintenance plan and historical efficiency; the order of operations and resource matching relationship are determined according to the process route of the order; the equipment maintenance plan is solidified into the resource unavailability period, and raw material supply delay constraint is generated by combining inventory, logistics in transit and procurement delivery cycle; the upper limit of energy is set by energy threshold / contract electricity capacity, the data statistics period and metering caliber are unified, and the initial production scheduling constraint conditions are generated.

[0081] It should be noted that the process route is obtained by extracting standard process documents from the manufacturing execution management information source and combining them with the parsing of the order product structure table.

[0082] The sources of equipment maintenance plans include pre-set periodic maintenance plans, estimated repair times extracted from abnormal maintenance records, and predictions of the next maintenance time window from historical maintenance cycle statistical models.

[0083] The sources of raw material supply delays include procurement cycle data recorded in the ERP system, the statistical average of historical supplier delivery times, and material shortage warnings and replenishment cycles in the warehousing system.

[0084] The sources of energy caps include energy dispatch thresholds for different time periods provided by energy monitoring platforms, factory contract power capacity or seasonal power rationing policies, and equipment energy consumption distribution estimated by historical energy consumption models.

[0085] Based on the initial production scheduling constraints, a digital twin model is constructed by combining the virtual production line topology, processing nodes, resource nodes, and logistics paths.

[0086] Furthermore, based on the initial production scheduling constraints, and according to the process route and resource deployment information, task dependencies and resource distribution are extracted to generate a virtual production line topology; combined with the effective status of the equipment maintenance plan, a set of resource nodes is determined from the resource capacity limit list; the order process is decomposed according to the process route, and a set of processing nodes is constructed by combining the accessibility conditions of raw material supply delay; using the resource node connections in the virtual production line topology as the skeleton, and combining the raw material supply delay and energy limit, the transmission capacity and constraints are verified to determine the set of logistics paths; and a digital twin model with a one-to-one correspondence between the virtual production line and the physical production line is established.

[0087] It should be noted that the resource node set corresponds to physical equipment, warehousing, and logistics hubs, and is determined based on the complete list of production and logistics resources listed in the resource capacity constraints, combined with the effective status screening results in the equipment maintenance plan.

[0088] Furthermore, through data mapping functions Physical production line at any time The state vector is mapped to the virtual production line state vector, represented as:

[0089] ;

[0090] in, Indicates the physical production line at time... The state vector, Indicates the virtual production line at time. The state vector, It represents a moment on a unified timeline.

[0091] Physical production line at all times The state vector is constructed from data collected on-site and continuously collected as production is executed, forming a sequence of state vectors of the physical production line that evolves over time.

[0092] It should be noted that the mapping function The physical production line state vectors at the same time are time-aligned and standardized, resource / node identifiers are encoded, and aggregated according to the virtual production line topology to generate the virtual production line at time [time value missing]. The state vector.

[0093] Furthermore, by using data mapping functions to achieve bidirectional synchronization between physical and virtual states, each element of the virtual production line has a clear semantic mapping in the virtual production line topology, resource nodes, processing nodes, and logistics paths. Combined with resource capacity constraints and equipment maintenance plans in the initial production scheduling constraints, the actual production state is mapped into a complete constraint state vector that can be used for simulation and scheduling optimization in the digital twin model.

[0094] S2. Predict the available capacity of production resources through digital twin models, identify risk periods, and update the twin status of production resources.

[0095] Based on the digital twin model, the available production capacity of production resources is predicted by combining historical operation data, maintenance plan data and logistics bottleneck data, and the prediction results of available production capacity are generated.

[0096] It should be noted that the historical operation data of production resources, maintenance plan data, and logistics bottleneck data are all uniformly measured in units of pieces per hour, with time variables measured in hours; all data are standardized before entering the digital twin model to ensure consistency in the dimensions of the integral items.

[0097] Furthermore, within the established digital twin model, the predicted available production capacity is generated using a formula based on historical production resource operation data, maintenance plan data, and logistics bottleneck data. Even further, the predicted average available production capacity is expressed as:

[0098] ;

[0099] in, This indicates the resources within the time interval [t0, t0+Δt]. The projected average available production resources capacity Indicates time The theoretical maximum output, Representing resources At any moment Production capacity loss due to planned maintenance Representing resources Production capacity losses caused by logistics delays The mean time between failures (MTBF) of the equipment is represented by λ, where λ represents the failure impact attenuation coefficient. Indicates the length of the prediction time window. This indicates that the starting point of the prediction time window is greater than 0. Represents the integral variable. ∈[t0,t0+Δt].

[0100] It should be noted that the time The theoretical maximum output is calculated by combining the resource capacity limit in the initial scheduling constraints with the virtual production line topology and the rated capacity of resource nodes at different times.

[0101] resource At any moment Production capacity loss due to planned maintenance is estimated based on equipment maintenance plans, historical maintenance records, and equipment operating parameters to determine the loss window and magnitude caused by maintenance. This indicates the planned maintenance of the corresponding capacity loss item.

[0102] resource The capacity loss caused by logistics delays is calculated by referring to the status of logistics nodes, energy consumption data, and logistics in-transit vectors, combined with the real-time congestion status of each logistics route, and the index is used. This represents the capacity loss item corresponding to logistics delays. The mean time between failures (MTBF) of the equipment is derived statistically from historical equipment operating data and maintenance plans. The failure impact attenuation coefficient is calibrated by minimizing the deviation between the actual available capacity in the historical period and the model's predicted value, and is updated on a rolling basis according to a sliding time window.

[0103] It should be noted that the integral variable It refers to the time points that slide within a short time interval [t0, t0+Δt] in the future, and is used to accumulate and weight the changes in the average available production capacity of production resources within the future time window.

[0104] The capacity loss caused by logistics delays is calculated by referring to the status of logistics nodes, energy consumption data, and logistics in-transit vectors, combined with the real-time congestion status of each logistics path.

[0105] The predicted average available production capacity takes into account multiple variables such as resource capacity, equipment maintenance, logistics bottlenecks, and resource reliability, enabling a dynamic quantitative expression of the average available production capacity and improving the scientific nature and responsiveness of task allocation and scheduling plans.

[0106] Furthermore, the integral calculation of the average available production capacity forecast adopts a sliding window method, and the forecast results are output in hourly increments to form a set of average available production capacity forecast results.

[0107] It should be noted that the set of average available production resources forecast results is a set formed by dynamically estimating the average available production resources for the next n forecast time windows.

[0108] Based on the average production resource available capacity forecast, time periods below the capacity threshold are identified as risk periods and the twin status of production resources is updated.

[0109] Furthermore, a capacity threshold is set. When the average available production capacity of the production resources corresponding to a time node is less than the capacity threshold, the time interval in which the time node is located is recorded as a risk period.

[0110] Among them, the capacity threshold is a benchmark value used to determine whether the capacity is in a risky state. It is calculated by statistically analyzing the mean and standard deviation of the actual historical available production capacity data and combining it with the set fault tolerance coefficient.

[0111] Based on the risk period, a set of risk periods is generated, and the twin state of production resources is updated.

[0112] Furthermore, the production resource status within the corresponding risk period is updated in the digital twin model to form a new set of production resource twin states.

[0113] In the digital twin model, status patches are written to the corresponding resource nodes at the granularity of "resource × risk period" to generate a new set of production resource twin states. Each status patch contains at least a derated operation flag and a capacity limit flag. The generated new set of production resource twin states can be regarded as a set of resource status records with an expiration date (such as: resource ID, time window start and end, derated operation flag, capacity limit value, etc.), which are directly read and executed in subsequent production scheduling to limit the allocation and output of resources within the corresponding time window.

[0114] It should be noted that when generating a new production resource twin state, some production resource available capacity limit marker parameters for scheduling optimization are added to the original physical state. These parameters are derived from the average production resource available capacity prediction results. First, the average production resource available capacity prediction result within each "resource × prediction time window" is calculated. Then, a capacity threshold is set based on historical data statistics to determine whether the time window belongs to a risk period. If it belongs to a risk period, the average production resource available capacity prediction result is used as the benchmark value of the maximum allowable available capacity of the resource within the time window, and it is reduced by a preset fault tolerance coefficient or the smaller value of the rated capacity is taken to generate the production resource available capacity limit marker parameters. At the same time, the derated operation flag is set.

[0115] It should be noted that the derated operation flag indicates that production resources are not operating at full capacity during certain forecast periods, but can only operate at lower efficiency due to factors such as maintenance, malfunctions, or bottlenecks.

[0116] The available capacity limit marker for production resources indicates the maximum permissible output capacity of production resources at certain times, derived from the average available capacity forecast of production resources.

[0117] The twin state during the risk period not only reflects the current available production capacity constraints, but also directly participates in the scheduling optimization calculation as a constraint in the subsequent scheduling phase, thereby realizing a prediction-driven production scheduling strategy.

[0118] S3. Couple the predicted available production capacity with the order set issued by the production management link, calculate the overall priority of the orders and generate a task priority queue, solidify the risk window and update the initial production scheduling constraints.

[0119] The production resource availability forecast is coupled with the order set issued by the production management process to generate the production resource and order set coupling result.

[0120] Furthermore, each order in the order set issued by the production management process includes the delivery time. Order quantity, customer rating, and required process resource path.

[0121] The order set issued in the production management process is mapped and matched with the average available production capacity forecast set in the time and resource dimensions to form a coupled structure of production resources and order set, represented as follows:

[0122] ;

[0123] in, This represents the set of results resulting from the coupling of production resources and orders. Indicates the first One order, Indicates order The corresponding set of production time requirements, This indicates that the traversal belongs to the order. All production time points , This indicates the resources within the time interval [t1, t1+Δt]. The projected average available production resources capacity.

[0124] Based on the coupling result of the average production resource availability forecast and the order set, the overall order priority of each order is calculated, and the overall order priority of all orders is generated.

[0125] When generating a task priority queue, a cross-time-domain sliding window mechanism is adopted to take into account not only the current capacity forecast, but also the resource supply and demand matching situation in multiple future time periods.

[0126] Furthermore, starting from the current moment, multiple time windows are generated according to the preset window length and step size, with each window covering the supply and demand of production resources for a future period of time.

[0127] Within each time window, based on the average available production capacity prediction results of the digital twin model, the available capacity values ​​of all average available production resources within the time window are statistically analyzed to form a window-level set of average available production resources.

[0128] Furthermore, based on the order delivery deadline, the order demand is distributed across relevant time windows to generate the nominal order demand in different time windows.

[0129] Within each time window, the average available production capacity and nominal order demand at the window level are compared to obtain the sufficiency of resource supply and demand for orders placed within the time window.

[0130] It should be noted that if the average available production capacity at the window level is insufficient, a supply-demand gap will be generated.

[0131] Furthermore, the supply and demand gaps across multiple time windows are weighted and accumulated, and cross-time-domain risk penalties are applied. To highlight the importance of adjacent windows, a decaying weight is used to distinguish between near-term and far-term windows, as shown below:

[0132] ;

[0133] ;

[0134] in, Indicates cross-time domain risk penalty points, Indicates the number of scrolling time windows. This represents the time window weighting function. Display window Risk gap items, Indicates order In the time window Resource sufficiency Indicates the time window index. This represents the decay coefficient, which controls the rate at which the weight decreases with the window index.

[0135] It should be noted that when the resource supply and demand sufficiency is less than 1, the risk gap term equals the gap ratio. When the resource sufficiency is greater than or equal to 1, the risk gap item is set to 0 and no penalty points are deducted.

[0136] The decay coefficient determines the decreasing rate by which closer windows have a larger weight and farther windows have a smaller weight. It is set based on the comprehensive rolling view strategy, the decrease in prediction confidence over time, window configuration, and business KPI preferences, and its value range is greater than 0.

[0137] Based on the urgency of order delivery and customer level, combined with cross-time-domain risk penalty points, the overall priority of the order is calculated.

[0138] Furthermore, the urgency of order delivery is calculated and normalized according to a unified standard, customer level is mapped to customer level coefficient, cross-time domain risk penalty is calculated by weighted accumulation according to time window, and the urgency of delivery and customer level coefficient are combined into a base score according to the configuration weight. The cross-time domain risk penalty is then used for correction to generate a comprehensive order priority.

[0139] Orders are sorted from highest to lowest overall priority to generate a task priority queue.

[0140] Furthermore, all orders are sorted from highest to lowest according to their overall priority, generating a cross-time-domain task priority queue.

[0141] It should be noted that the priority queue across time domains is recalculated each time the time rolling window is updated, so that the scheduling strategy takes into account both short-term and long-term needs.

[0142] Based on the task priority queue and risk window, the risk window is fixed and the initial production scheduling constraints are updated.

[0143] Furthermore, based on the identified risk period set, the task priority queue and risk window under the current time round are jointly mapped, and it is analyzed whether there is risk coverage in the expected processing cycle of each task. This risk information is written into the scheduling constraint structure, and finally the updated production scheduling constraints are formed.

[0144] S4. Based on the task priority queue and the updated production scheduling constraints, perform scheduling optimization in the digital twin model to generate candidate scheduling schemes, and perform time-series simulation to detect conflicts on the candidate scheduling schemes in the digital twin environment.

[0145] Based on the task priority queue and the updated production scheduling constraints, scheduling optimization operations are performed in the digital twin model to generate a set of scheduling optimization operations.

[0146] Furthermore, the scheduling optimization algorithm is invoked in the digital twin model. Taking the task priority queue and the updated production scheduling constraints as input, the algorithm makes optimization decisions on task allocation, start time and processing resources in terms of time, resource and path dimensions, and outputs multiple scheduling results, which consist of multiple candidate scheduling schemes.

[0147] It should be noted that each candidate scheduling scheme includes a task identifier, the resource number to which the time is allocated, the task start time, and the task end time.

[0148] It should be noted that the multiple output scheduling results are scheduling schemes that satisfy the updated production scheduling constraints, generated by inputting the task priority queue and the updated production scheduling constraints into the digital twin model and using a multi-iteration optimization algorithm.

[0149] It should be noted that the candidate scheduling schemes include a three-dimensional mapping structure of task-resource-time.

[0150] The task identifier is a unique identifier for a specific task during the scheduling process. In the scheduling scheme, the task is uniquely indexed and bound to the scheduling information.

[0151] In a digital twin environment, timing simulations are performed on the scheduling optimization computation set to generate simulation results for candidate scheduling schemes.

[0152] Furthermore, the timing simulation process considers the sequential dependencies between tasks, the exclusivity of resource usage, and the time consumption of logistics transportation. It performs scheduling operations on a virtual time axis and outputs a set of simulation results for candidate scheduling schemes.

[0153] Deploy discrete event simulations in the completed digital twin model, using event queues to drive time progression; load the initial production state, assemble updated production scheduling constraints and task priority queues, and inject the current average available production resources forecast as resource capacity parameters within the simulation time window.

[0154] Each scheduling scheme in the scheduling optimization operation set is input into the simulation as a "dispatch and allocation strategy": when a task meets the prerequisite completion conditions and enters the executable queue, the target resource allocation and logistics path are selected according to the process route, resource locking and occupation duration calculation are performed, start / end events are generated and written to the event queue.

[0155] At key points such as task initiation, resource application, and logistics dispatch, three types of conflicts are detected in real time according to rules: verifying job sequence dependencies, identifying processing order conflicts, verifying mutual exclusion of resource occupancy within the same time window, identifying resource competition conflicts, and calculating delay exceedances and identifying logistics congestion conflicts based on path remaining capacity and in-transit queues.

[0156] For triggered conflicts, the conflict location, time point, involved task and resource identifiers, and conflict type are recorded and written to the simulation log of the current scheme. During the simulation process, a simulation timeline, resource usage records, logistics path delays, and task status markers are generated synchronously. After the simulation ends, the scheduling execution trajectory and conflict list of the candidate scheduling schemes are output, generating a complete simulation result of the candidate scheduling schemes. The process is repeated to obtain the simulation results of all candidate scheduling schemes, and the results are summarized to form a set of simulation results of the candidate scheduling schemes.

[0157] The simulation results of the candidate scheduling schemes detected three types of conflicts, including processing sequence conflicts, resource competition conflicts, and logistics congestion conflicts.

[0158] It should be noted that processing sequence conflict is when the task execution order is inconsistent with the process route; resource competition conflict is when the same resource is occupied by multiple tasks in the same time period; logistics congestion conflict is when the logistics path capacity is over-utilized or the path delay exceeds the limit.

[0159] Each conflict detection result includes the conflict type, the task identifier involved in the conflict, the resource identifier where the conflict occurred, and the time and location of the conflict.

[0160] During the simulation, the execution time, resource occupancy status, and logistics path of each task are tracked in real time to identify potential processing sequence conflicts, resource competition conflicts, and logistics blockage conflicts. By traversing the simulation trajectory, information such as the task identifier, allocated resource number, conflict type, and conflict time for each type of conflict is recorded to form conflict information under a single scheme. Based on this method, the same operation is performed on all candidate scheduling schemes to finally construct a complete set of conflict detection results.

[0161] S5. Based on the detected conflicts, write back the conflict information to the updated production scheduling constraints, perform local rearrangement within the same round until the conflicts are eliminated and the performance indicators meet the preset requirements, and output an executable scheduling scheme.

[0162] A set of conflict information is generated based on the set of conflict detection results.

[0163] The overall conflict information set is composed of each individual conflict information data.

[0164] Furthermore, starting with each individual conflict information data, similar items are first merged and field naming and time order are standardized to generate a conflict detection result set. The task identifier, resource identifier, conflict time and conflict type are extracted one by one, and they are arranged into standard items in a fixed order. Deduplication, missing data are filled in and consistency is checked, and finally the conflict information set is output.

[0165] Conflict information data includes conflict task identifier, conflict resource identifier, conflict time, and conflict type.

[0166] Write back the conflict information to the updated production scheduling constraints, and generate the updated scheduling constraint results including the conflict constraints.

[0167] Furthermore, the conflict information set is written back to the updated production scheduling constraints to form new conflict constraints, generating updated production scheduling constraints that include conflict constraints.

[0168] The updated production scheduling constraints, including conflict constraints, are generated by merging the newly added conflict constraints and the updated production scheduling constraints.

[0169] It should be noted that the newly added conflict constraints are used to avoid resource usage conflicts, incorrect processing sequences, and logistical blockages.

[0170] Furthermore, the conflict detection results, including conflict task identifiers, resource identifiers, conflict times, and conflict types, need to be extracted in a structured manner to generate new conflict constraints. These new constraints explicitly restrict certain resources from being occupied by conflict tasks within a specific time period, prohibit the execution of tasks that violate the established processing order, or restrict task triggering under logistics congestion conditions, thereby enabling the control and avoidance of conflict scenarios. The updated production scheduling constraints and the new conflict constraints are then merged to generate updated production scheduling constraints that include conflict constraints.

[0171] Based on the updated production scheduling constraints, including conflict constraints, a local rearrangement operation is performed within the same round to generate the local rearrangement result.

[0172] Furthermore, based on the updated production scheduling constraints, including conflict constraints, a local rearrangement operation is performed within the same round. The affected task subset is rearranged using a local search algorithm to generate a set of local rearrangement schemes.

[0173] It should be noted that the set of local rearrangement schemes consists of multiple records, each of which includes a task identifier, a rearranged resource identifier, a rearranged task start time, and a rearranged task end time.

[0174] Furthermore, based on the set of local rearrangement schemes, and taking the original scheduling scheme as a benchmark, conflicting task records are removed, and the corresponding rearrangement records in the set of local rearrangement schemes are used to replace or supplement them to generate an updated scheduling scheme.

[0175] Furthermore, starting with the set of results from the local reordering operation, including the reordering resources and start and end times of the affected tasks, the old arrangements of the corresponding conflicting tasks are located and removed from the original scheduling scheme. Using the empty spaces after removal as input, the new arrangements from the set of results from the local reordering operation are inserted and placed in place according to time order and resource number. Using the inserted temporary scheme as input, sequence verification, resource occupancy verification, and logistics verification are performed based on the updated production scheduling constraints, including conflict constraints. If any violations are found, new constraints are generated and the local search is returned to continue reordering. Using the verified temporary scheme as input, the existing arrangements of unaffected tasks are merged to form the locally reordered scheduling scheme, which is then entered into simulation verification.

[0176] Based on the results of the local rearrangement operation, the timing simulation is executed again to repeatedly check whether the conflict is eliminated and whether the performance indicators meet the preset requirements. When the conditions are met, an executable scheduling scheme is generated.

[0177] Furthermore, the scheduling scheme after local rearrangement is simulated to detect conflicts and perform performance evaluation. If the conflict detection result is empty and the performance evaluation meets the preset performance index threshold, the scheduling scheme after local rearrangement is output as the final executable scheduling scheme.

[0178] Furthermore, using the partially rearranged scheduling scheme as input, the execution process is reviewed in the digital twin environment, and the simulation timeline, resource usage records, and logistics path delays are summarized to generate simulation results for performance evaluation. The overall performance is calculated based on indicators such as delivery date achievement rate, total project duration, resource load balancing, logistics achievement rate, and energy consumption, forming the performance evaluation results. These results are then compared item by item with pre-set performance indicator thresholds to determine whether the partially rearranged scheduling scheme passes the evaluation.

[0179] The completion time of each order is determined based on the simulation timeline. Comparing this timeline with the delivery deadline, the percentage of orders completed on time is calculated to generate the delivery achievement rate. The total project duration is calculated based on the span between the earliest start time and the latest finish time of all tasks. Based on resource usage records, the average utilization rate of each resource within a window is calculated, and resource load balancing is measured by the volatility (such as variance or coefficient of variation) of each resource utilization rate; lower volatility results in a higher score. Based on logistics path delays and in-transit records, the percentage of logistics events arriving within the planned arrival or tolerance threshold is calculated to generate the logistics achievement rate. Energy consumption is calculated by combining resource operation / idle periods and power consumption coefficients.

[0180] For indicators such as delivery date achievement rate, total project duration, resource load balancing, logistics achievement rate, and energy consumption, normalization is performed using the same caliber: for indicators where a larger value is better (delivery date achievement rate, logistics achievement rate), a direct mapping is performed to a score; for indicators where a smaller value is better (total project duration, energy consumption, load imbalance), a reverse mapping is performed or a score is converted according to a threshold. A comprehensive performance is generated by weighting the configured indicators and comparing them item by item with the preset performance indicator thresholds to determine whether the local rescheduling plan is approved.

[0181] If all indicators reach the preset performance threshold and the conflict detection is empty, the scheduling scheme after local rearrangement is confirmed as executable and output.

[0182] Pre-set performance index thresholds can be determined by statistically analyzing the mean and standard deviation of relevant historical data and combining them with a fault tolerance coefficient.

[0183] If the simulation results of the scheduling scheme after local rearrangement do not meet the requirement of empty conflict detection results, or the performance evaluation results do not meet the preset performance index threshold, the following steps are executed in sequence: simulation evaluation, generating a conflict information set, writing back and updating scheduling constraints, identifying affected tasks and performing local rearrangement, updating the scheduling scheme and evaluating again. The above steps are repeated until the conflict is eliminated and the performance index reaches the preset performance index threshold.

[0184] It should be noted that by dynamically feeding back and iteratively updating production scheduling constraints through conflict information, the adaptability and real-time response capability of scheduling results are improved, ensuring that key conflicts are avoided while maintaining scheduling efficiency, and achieving the stability and executability of the final scheduling scheme.

[0185] S6. Distribute the executable scheduling plan to the production and logistics links, monitor execution deviations, generate local work condition packages, reschedule affected tasks, complete the overall production task, write the production data back to the digital twin database, and update the production resource available capacity forecast data, production scheduling constraints, and optimization weights.

[0186] The executable scheduling plan is distributed to the production and logistics stages, and the execution deviation during the execution of the executable scheduling plan is monitored to generate execution deviation results.

[0187] Furthermore, the generated executable scheduling plan is distributed to the actual production and logistics links, a monitoring cycle is set, and status data is periodically collected and compared with the scheduling plan value to form an execution deviation set.

[0188] The execution deviation set consists of multiple records. Each record includes the task in which the deviation occurred, the deviation type, and the deviation time. The deviation time is the difference between the actual time and the planned time. The deviation types include start delay, logistics interruption, and resource unavailability.

[0189] Based on the set of execution deviations, a local condition package is constructed for all affected tasks.

[0190] It should be noted that a local condition package consists of multiple records, each of which includes the scope of the task in which the deviation occurred, the type of deviation, and the duration of the deviation.

[0191] Furthermore, based on the execution deviation results fed back from the production or logistics stages, the specific task in which the deviation occurred is identified; the deviation propagation analysis model is invoked, and the scheduling relationship, resource usage information, and dependency chain of the current task are input to dynamically calculate the potential impact range of the deviation; the deviation type (such as delay, resource unavailability, path blockage, etc.) and deviation time are recorded according to the collected deviation information and assembled into a structure to form a local task condition package instance. All deviation tasks are identified, propagated, recorded, and local task condition packages are assembled one by one, ultimately forming a set of local task condition packages to support subsequent execution rescheduling.

[0192] It should be noted that the process of constructing the deviation propagation analysis model includes determining the inputs and outputs. The inputs include the scheduling relationship of tasks, resource usage information, material / BOM dependency chain, and deviation events observed in the digital twin model. The outputs are the set of affected tasks, resource overload time window, and risk assessment of critical path / delivery nodes, and record the deviation type, deviation time, affected task identifier, and impact scope.

[0193] Using tasks as nodes, establish multiple types of edges according to process sequence, shared resource relationships, and material / BOM chains; record the earliest / latest start and completion times of tasks on task nodes, and record resource capacity and occupancy information on the resource side to construct a task-resource multi-relationship graph.

[0194] Deviation events observed in the digital twin model are normalized as node equivalent disturbances (changes in schedule / quality rework / resource usage), and the disturbed task nodes and time windows are located on the task-resource multi-relationship graph.

[0195] Three types of propagation cores are defined: the earliest / latest moment of the update of the process pre- and post-process relationship affecting the subsequent task; the resource relationship detects resource capacity overload and performs local rearrangement of conflicting tasks, so that the delay is propagated between tasks with the same resource; and the material / BOM chain transmits the change of start-up time to downstream tasks along the material dependency.

[0196] Add the disturbed task nodes to the event queue and advance the update of the three types of propagation kernels according to the topology direction; after each update, add the newly affected tasks back to the event queue. When the earliest / latest start and finish times of all tasks are consistent with the previous round after one propagation and local resource rearrangement, no new resource capacity overload time window is generated, no further local rearrangement is needed, and the event queue is empty, stop the iteration.

[0197] Output the affected task identifiers and the start, completion, and margin changes of each task. Output the scope of impact, including the set of affected tasks, the capacity overload time windows of each resource, and the changes and worst-case delays of the critical path and delivery nodes. Record the deviation type and deviation time and write them into the local condition package.

[0198] The local condition package includes the affected task identifier, the scope of impact, the type of deviation, and the time of deviation.

[0199] Based on the local condition package set, a rescheduling operation is performed to generate the rescheduling operation result.

[0200] Furthermore, based on the local condition package set, the affected task set is extracted, represented as:

[0201] ;

[0202] in, This represents the set of affected tasks extracted from the local condition package. Represents a set of local operating condition packages. Indicates to Perform a union operation on all records. Indicates the specific task in which the deviation occurred, using the subscript. This represents the index used to perform a union operation on each record in the local condition package set. This represents the set of tasks affected by the deviation. Indicates the type of deviation. Indicates the duration of the deviation.

[0203] Furthermore, in order to extract all tasks that need to be rescheduled, it is necessary to perform a set-and-merge operation on the tasks involved in all local task packages.

[0204] Specifically, the set of local work condition packages is traversed, the task that deviated is combined with the scope of influence of the task that deviated, and all the union results are merged again to form a set of tasks in a local work condition package, which provides accurate task boundaries for subsequent construction of local rescheduling constraints and rescheduling calculations.

[0205] Furthermore, local rescheduling constraints are constructed, and a rescheduling algorithm is executed based on the current remaining resource status to form a rescheduling scheme.

[0206] The rescheduling scheme consists of multiple records, each of which includes the rescheduled task, the new resources allocated to the task, the new start time of the task, and the new end time of the task.

[0207] Furthermore, based on the latest scheduling execution status, remaining production resources, and current scheduling constraints, local rescheduling constraints are first constructed; then, a rescheduling algorithm is used to efficiently rearrange the tasks in the task set within the local task package.

[0208] It should be noted that the efficient rescheduling method is "time window sliding + greedy insertion": For each task affected by rescheduling, its expected time window is slid, and an insertion gap is prioritized in the new resources allocated to the task in rescheduling; under the premise of satisfying the timing constraints and resource constraints, the start and end times of the tasks affected by rescheduling are arranged as early as possible; finally, the scheduling information of all successfully rescheduled tasks is recorded in a unified manner to form a complete set of rescheduling schemes, thereby realizing dynamic scheduling adjustment without interrupting tasks.

[0209] The rescheduling scheme is merged into the remaining unaffected scheduling plans to form a complete update scheme, represented as:

[0210] ;

[0211] in, This represents the latest executable task scheduling scheme generated after rescheduling. This represents the new set of task schedules generated by the rescheduling. This represents the set of scheduling schemes currently being executed. This represents the set of affected tasks extracted from the local condition package.

[0212] Furthermore, firstly, based on the set of affected tasks identified by the local task package, task scheduling records are removed from the currently executing scheduling scheme to eliminate disturbed and unstable plans. Then, a union operation is performed on the newly generated task scheduling set and the removed currently executing scheduling scheme to achieve seamless replacement of task schedules. During the merging process, it must be ensured that the resource allocation and timing of each task in the newly generated task scheduling set do not conflict with the remaining tasks in the currently executing scheduling scheme set, thereby constructing a complete update scheme that combines consistency and executability.

[0213] Based on the rescheduling operation results, the overall production task is completed and the production task completion result is generated.

[0214] Furthermore, the latest executable task scheduling scheme generated after rescheduling is fully executed to the final state, and the task completion status, start and completion time, resource usage records, etc. are recorded to form the production task completion result.

[0215] Based on the results of production task completion, production data is written back to the digital twin database to update the production resource availability capacity forecast data, production scheduling constraints, and optimization weights, thereby generating an updated digital twin database.

[0216] Furthermore, the results of production task completion are written back to the digital twin database to update the production resource availability forecast data, represented as:

[0217] ;

[0218] in, Representing resources At any moment The updated historical usage records include the latest task execution status. Representing resources At any moment The original historical occupancy records, that is, the set of occupancy information that existed before the update. A occupancy record triple represents a resource. At the actual start time of the task until the actual end of execution time Occupancy status, This indicates a specific production resource, such as a machine tool, a production line, or a logistics vehicle. The actual start time of the task The actual completion time of the task.

[0219] Furthermore, the actual execution time of the resources involved in each completed task is determined by the actual start and end times, forming a new resource occupancy record triplet. This new resource occupancy record triplet will be compared with the current resource occupancy record at time... The original historical occupancy records are merged to form the updated resource occupancy history. The update process ensures the continuity and real-time nature of resource status information, providing more accurate basic data for subsequent prediction of available production capacity, thereby improving the predictive model's responsiveness to resource bottleneck identification and scheduling strategy adjustment.

[0220] Furthermore, update production scheduling constraints, such as dynamic energy consumption constraints, logistics delays, and resource maintenance cycles.

[0221] The set of constraints for the current round is combined with the execution results of the current round to generate the set of scheduling constraints for the next round.

[0222] It should be noted that the results of this round of production task completion include information such as task execution status, actual resource usage, and task start and end times.

[0223] The set of constraints in this round is the final set of scheduling constraints that includes optimization weights, conflict constraints, and feasibility rules.

[0224] Furthermore, based on resource usage records during task execution, parameters such as resource maintenance cycles, available time windows, and energy consumption boundaries are updated to reflect the latest equipment health status and energy consumption. Combined with logistics execution paths and completion times, the latency prediction rules and availability windows for logistics node status are dynamically adjusted to strengthen the constraint on material flow bottlenecks.

[0225] If resource restructuring or plan adjustments are detected during task execution, the process path and resource alternatives also need to be updated.

[0226] By comprehensively applying the task status, actual execution time, and conflict feedback involved in the completion results of the current round of production tasks, and transforming the set of production scheduling constraints from the previous round through a function, new production scheduling constraints are generated, including conflict avoidance, resource updates, energy consumption limits, logistics adjustments, and optimization weights, providing constraint support for the next round of scheduling optimization.

[0227] In a preferred embodiment, the process of updating the optimization weights not only corrects a single optimization objective based on execution deviations, but also adopts a dynamic weight adaptive evolution mechanism for optimization objectives, so that the weights of each optimization objective can evolve round by round with scheduling performance.

[0228] Furthermore, after each round of scheduling is completed, multiple types of indicator data are collected to generate a scheduling performance set for the current round.

[0229] It should be noted that the various data indicators include delivery delay duration, overall project duration, energy consumption, on-time logistics rate, and resource load balance.

[0230] Furthermore, the data of various indicators are normalized and standardized using a sliding statistics method to eliminate the dimensional differences between different indicators and obtain the relative deviation values ​​of various indicators.

[0231] It should be noted that the relative deviation values ​​of multiple indicators can reflect the fluctuation of each indicator relative to its historical average level.

[0232] A comprehensive driving signal is constructed based on the relative deviation values ​​of multiple indicators, and a weight update vector is generated.

[0233] Furthermore, a driving vector consisting of the relative deviations of multiple indicators is generated. The weight update amount is calculated by optimizing the target weight and the driving vector, and is expressed as:

[0234] ;

[0235] in, Indicate the optimization objective In the currently ongoing [number] The weight update amount of the round, Indicates the first The original target weights during round-robin scheduling Indicates the first The normalized index deviation of the optimization objective of the wheel, This represents the sensitivity coefficient.

[0236] It should be noted that the sensitivity coefficient controls the degree of influence of the deviation signal on the weight update amplitude, which is set based on the dynamics and stability of the production environment and the simulation scheduling results. The example value range is (0.05, 1). Values ​​that are too small (e.g.) This can lead to almost no change in weight updates, losing its adaptive meaning, and excessively large values ​​(e.g.) This can lead to overly drastic weight updates and unstable scheduling results.

[0237] Furthermore, the weight update amounts of all optimization objectives are normalized, and a minimum weight threshold is set for each optimization objective to ensure that no optimization objective is completely ignored.

[0238] It should be noted that the final normalized weights satisfy the condition that the sum is 1 and none are less than the preset minimum weight threshold, expressed as:

[0239] ;

[0240] in, Indicates the first During the round of iteration, the first The final normalized weights for each optimization objective, Indicates the first During the round of iteration, the first The initial weight values ​​for each optimization objective. Indicates the first In each iteration, the target in the target set is optimized. The initial weight value, index This represents the target index of the target set to be optimized. This represents the preset minimum weight threshold.

[0241] The new weight vector is used in the scheduling optimization objective function of the next round, so that the scheduling optimization can adaptively focus on the optimization objective that performed poorly in the previous round. The scheduling optimization objective can be continuously adjusted between different rounds, thus forming a continuously evolving dynamic weight mechanism.

[0242] This embodiment also provides a production scheduling device based on digital twins, including:

[0243] The data twin building module is used to collect production data, generate the initial production state, define the initial production scheduling constraints, and build a digital twin model.

[0244] The capacity forecasting and risk control module is used to forecast the available capacity of production resources, identify risky periods, and update the twin status of production resources.

[0245] The order priority queue module is used to couple the predicted available production capacity with the order set issued by the production management process, calculate the overall priority of orders and generate a task priority queue, solidify the risk window and update the initial production scheduling constraints.

[0246] The simulation conflict detection module is optimized. Based on the task priority queue and the updated production scheduling constraints, scheduling optimization is performed in the digital twin model to generate candidate scheduling schemes. The candidate scheduling schemes are then subjected to time-series simulation to detect conflicts in the digital twin environment.

[0247] The conflict write-back and reordering module is used to write back conflict information to the updated production scheduling constraints, perform local reordering within the same round until the conflict is eliminated and the performance indicators meet the preset requirements, and output an executable scheduling scheme.

[0248] The execution monitoring and rescheduling module is used to distribute executable scheduling plans to the production and logistics links, monitor execution deviations, generate local work condition packages, reschedule affected tasks, complete the overall production tasks, write production data back to the digital twin database, and update production resource availability capacity forecast data, production scheduling constraints, and optimization weights.

[0249] This embodiment also provides a computer device applicable to the production scheduling method based on digital twins, including: a memory and a processor, wherein the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the production scheduling method based on digital twins as proposed in the above embodiment.

[0250] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0251] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the production scheduling method based on digital twins as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0252] In summary, this invention, by predicting the available capacity of production resources based on a digital twin model, identifying risk periods, and updating the twin state of production resources, achieves forward-looking perception of future capacity fluctuations before scheduling, thereby dynamically avoiding potential risks during the scheduling process and ensuring the continuity and reliability of resource allocation. By coupling the predicted available capacity of production resources with the order set issued by the production management link, calculating the comprehensive priority of orders and generating a task priority queue, solidifying the risk window and updating the initial production scheduling constraints, this invention achieves deep matching and ranking of order demand and resource capacity, followed by scheduling optimization and rearrangement, improving the accuracy and timeliness of scheduling.

[0253] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A production scheduling method based on digital twins, characterized in that: include, Collect production data, generate initial production status, define initial production scheduling constraints, and build a digital twin model; Predict the available capacity of production resources through digital twin models, identify risk periods, and update the twin status of production resources. Couple the predicted available production capacity with the order set issued by the production management link, calculate the overall priority of the orders and generate a task priority queue, solidify the risk window and update the initial production scheduling constraints; Based on the task priority queue and the updated production scheduling constraints, scheduling optimization is performed in the digital twin model to generate candidate scheduling schemes. The candidate scheduling schemes are then subjected to time-series simulation to detect conflicts in the digital twin environment. Based on the detected conflicts, write back the conflict information to the updated production scheduling constraints, perform local rearrangement within the same round until the conflicts are eliminated and the performance indicators meet the preset requirements, and output an executable scheduling scheme. The executable scheduling plan is distributed to the production and logistics links, execution deviations are monitored, local work condition packages are generated, affected tasks are rescheduled, the overall production task is completed, and production data is written back to the digital twin database to update the production resource available capacity forecast data, production scheduling constraints and optimization weights. Initial production scheduling constraints include resource capacity limits, process routes, equipment maintenance plans, raw material supply delays, and energy ceilings; Data sources for predicting the available capacity of production resources include historical operating data of production resources, equipment maintenance plan data, and logistics bottleneck data; The risk period refers to the time when production capacity is below the threshold. Solidifying the risk window and updating the initial production scheduling constraints refers to jointly mapping the task priority queue under the current time round with the risk window based on the identified risk period set, analyzing whether there is risk coverage in the expected processing cycle of each task, and writing the risk information into the scheduling constraint structure to finally form the updated production scheduling constraints. The types of conflicts detected include processing sequence conflicts, resource competition conflicts, and logistics congestion conflicts. Writing back conflict information to the updated production scheduling constraints means writing back conflict information to the updated production scheduling constraints to form new conflict constraints. This generates updated production scheduling constraints that include conflict constraints. The new conflict constraints are used to avoid resource competition conflicts, processing sequence conflicts, and logistics congestion conflicts.

2. The production scheduling method based on digital twins as described in claim 1, characterized in that: The construction of the digital twin model refers to collecting production data and aligning it with timestamps to generate a production dataset, generating an initial production state based on the production dataset, defining initial production scheduling constraints, and constructing a digital twin model by combining virtual production line topology, resource nodes, processing nodes, and logistics paths.

3. The production scheduling method based on digital twins as described in claim 2, characterized in that: The specific steps for predicting the available production capacity of production resources, identifying risk periods, and updating the twin status of production resources are as follows: Based on the digital twin model, the available capacity of production resources is predicted, and the prediction results of the available capacity of production resources are generated. Based on the predicted available production capacity of generated resources, time periods below the production capacity threshold are identified as risk periods, and the twin status of production resources is updated.

4. The production scheduling method based on digital twins as described in claim 3, characterized in that: The steps involve coupling the predicted available production capacity with the order set issued by the production management department, calculating the overall priority of the orders and generating a task priority queue, fixing the risk window, and updating the initial production scheduling constraints. The production resource availability forecast results are coupled with the order set issued by the production management process to generate the production resource and order set coupling result; Calculate the overall order priority for each order, and generate the overall order priority for all orders; Sort orders by overall priority from highest to lowest to generate a task priority queue; Solidify the risk window and update the initial production scheduling constraints.

5. The production scheduling method based on digital twins as described in claim 4, characterized in that: The execution scheduling optimization involves generating candidate scheduling schemes and performing time-series simulations to detect conflicts in the candidate scheduling schemes within a digital twin environment. The specific steps are as follows: Perform scheduling optimization operations in the digital twin model to generate a set of scheduling optimization operations; In a digital twin environment, a time-series simulation of the scheduling optimization computation set is performed to generate simulation results for candidate scheduling schemes; The simulation results of candidate scheduling schemes are used to detect processing sequence conflicts, resource competition conflicts, and logistics blockage conflicts, and conflict detection results are generated.

6. The production scheduling method based on digital twins as described in claim 5, characterized in that: The process involves writing back conflict information to the updated production scheduling constraints, performing local rearrangements within the same round, until the conflicts are resolved and performance indicators meet preset requirements, and then outputting an executable scheduling scheme. The specific steps are as follows: Based on the conflict detection results, the conflict type, conflict task identifier, conflict resource identifier, as well as the occurrence time and location are extracted and the conflict information is generated. Write back the conflict information to the updated production scheduling constraints, and generate the updated production scheduling constraints result including the conflict constraints; Based on the updated production scheduling constraint results of conflict constraints, perform local rearrangement operations within the same round to generate local rearrangement operation results; Based on the results of the local rearrangement operation, the timing simulation is executed again to repeatedly check whether the conflict is eliminated and whether the performance indicators meet the preset requirements. When the conditions are met, an executable scheduling scheme is generated.

7. The production scheduling method based on digital twins as described in claim 6, characterized in that: The process involves distributing the executable scheduling plan to the production and logistics stages, monitoring execution deviations, generating local work condition packages, rescheduling affected tasks, completing the overall production task, writing production data back to the digital twin database, and updating production resource availability forecast data, production scheduling constraints, and optimization weights. The specific steps are as follows: The executable scheduling plan is distributed to the production and logistics stages, the execution deviations during the execution of the executable scheduling plan are monitored, the execution deviation results are generated, and local work condition packages are generated based on the execution deviation results. Perform a rescheduling operation and generate the rescheduling result; Complete the overall production task and generate the production task completion result; Production data is written back to the digital twin database to update the production resource availability forecast data, production scheduling constraints, and optimization weights, thereby generating an updated digital twin database.

8. A production scheduling device based on digital twins, based on the production scheduling method based on digital twins according to any one of claims 1 to 7, characterized in that: include, The data twin building module is used to collect production data, generate the initial production state, define the initial production scheduling constraints, and build a digital twin model. The capacity forecasting and risk control module is used to forecast the available capacity of production resources, identify risky periods, and update the twin status of production resources. The order priority queue module is used to couple the predicted available production capacity with the order set issued by the production management link, calculate the overall priority of the orders and generate the task priority queue, solidify the risk window and update the initial production scheduling constraints. The simulation conflict detection module is optimized. Based on the task priority queue and the updated production scheduling constraints, scheduling optimization is performed in the digital twin model to generate candidate scheduling schemes. The candidate scheduling schemes are then subjected to time-series simulation to detect conflicts in the digital twin environment. The conflict writeback and reordering module is used to write back conflict information to the updated production scheduling constraints, perform local reordering in the same round until the conflict is eliminated and the performance indicators meet the preset requirements, and output an executable scheduling scheme. The execution monitoring and rescheduling module is used to distribute executable scheduling plans to the production and logistics links, monitor execution deviations, generate local work condition packages, reschedule affected tasks, complete the overall production tasks, write production data back to the digital twin database, and update production resource availability capacity forecast data, production scheduling constraints, and optimization weights.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the production scheduling method based on digital twins as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the production scheduling method based on digital twins as described in any one of claims 1 to 7.

Images