An online scheduling method for additive manufacturing considering load balancing

Abstract

This application provides an online scheduling method for additive manufacturing that considers load balancing, relating to the fields of additive manufacturing and intelligent manufacturing scheduling technology. The method includes: when a new order arrives, constructing a comprehensive priority model based on the order's static attributes and the system's dynamic state, and calculating a comprehensive priority score; determining the order's position in the queue using a three-stage dynamic insertion strategy based on the relationship between the priority score and a threshold; for large orders, splitting them into multiple devices based on an iterative greedy expected completion time balancing allocation algorithm; for orders that are not split, selecting the optimal device using a multi-objective optimization method based on an unused device index reward mechanism; and performing multi-level sorting of orders in the queue, scheduling them sequentially, and updating the system state. This invention can significantly reduce the maximum completion time in additive manufacturing scenarios with multiple product types, variable batches, and dynamically arriving orders, and is suitable for large-scale additive manufacturing production of complex functional components for aerospace.

Description

Technical Field

[0001] This application relates to the fields of additive manufacturing and intelligent manufacturing scheduling technology, and in particular to an online scheduling method for additive manufacturing that takes load balancing into account. Background Technology

[0002] With the maturation of additive manufacturing technology, especially driven by Industry 4.0 and the "Making as a Service (MaaS)" model, additive manufacturing has gradually evolved from simply producing complex parts to a key manufacturing method oriented towards on-demand customization and mass production. Additive manufacturing companies need to handle a large number of online orders simultaneously in a multi-factory, multi-equipment environment. These orders often have different delivery dates, priorities, and processing requirements, and the production system must respond in real time under limited capacity and tight scheduling conditions.

[0003] Against this backdrop, existing scheduling methods still generally suffer from the following shortcomings: First, most methods prioritize orders based on a single factor such as delivery date, arrival time, or processing time, lacking a comprehensive characterization of multi-dimensional static and dynamic characteristics such as order importance, waiting time, and equipment status; second, load balancing mechanisms are mostly static or only consider instantaneous load, lacking a systematic consideration of historical usage and the "unused equipment priority" strategy, which can easily lead to some equipment being overloaded for a long time while other equipment is idle; third, large orders are often split into simple fixed proportions or equal amounts, making it difficult to achieve fine-grained balanced allocation under multi-device parallel operation by combining expected completion time; fourth, the online response capability to uncertain events such as dynamic order arrival, equipment failure, and processing time fluctuations is limited, making it difficult to support the engineering requirements of rapid production changeover and real-time scheduling.

[0004] Therefore, there is an urgent need for an online scheduling method that can integrate multiple static and dynamic factors in additive manufacturing scenarios, while taking into account real-time response, intelligent order splitting, and long-term load balancing, in order to improve resource utilization, shorten overall completion time, and reduce production fluctuations under frequent production changeover conditions. Summary of the Invention

[0005] The purpose of this invention is to address the problems of untimely response, unbalanced load, and insufficient resource utilization in existing online scheduling methods for additive manufacturing, and to provide an online scheduling method that considers load balancing for additive manufacturing.

[0006] The above-mentioned objective of this application is achieved through the following technical solution: S1: Obtain the order parameters of the new order and the current status parameters of the additive manufacturing equipment; construct a comprehensive priority model and calculate the comprehensive priority score P of the new order; S2: Based on the relationship between the comprehensive priority score P and the multi-level priority threshold, new orders are divided into different priority levels, and a segmented dynamic insertion strategy is used to insert them into different positions in the online order queue. S3: For orders to be dispatched, calculate the comprehensive score from the set of candidate equipment that can process the corresponding products, select the equipment with the best comprehensive score as the target equipment, and generate the production task on the target equipment. S4: Sort the orders to be scheduled in the online order queue according to the preset multi-level sorting rules, execute production scheduling in sequence, and update the equipment availability time, equipment historical usage count and the current system time until the orders to be scheduled at the current moment are processed.

[0007] Optionally, step S1 includes: Order parameters must include at least one or more of the following: order arrival time, order quantity, basic priority, delivery date or delivery urgency, order value, and product importance; The equipment status parameters include at least one or more of the following: equipment availability time, equipment load, historical usage count, and processing efficiency; among which, equipment load is a status variable that characterizes the future occupancy level of the equipment. The comprehensive priority model is built based on multi-dimensional features, including static attribute features of orders and dynamic state features of the system; The overall priority score P is obtained by mapping the original overall score to the interval [0, 100] using a logarithmic compression normalization function.

[0008] Optionally, step S2 includes: S21: Based on the comprehensive priority score P and the preset priority threshold , , The relationship between them determines the priority level of new orders. ; among them when When is of extremely high priority, Time is of high priority, when When it is of medium priority, It is a low priority. S22: Based on priority levels, a three-stage dynamic insertion scheduling strategy is adopted to determine the insertion position of new orders in the online order queue; S23: Determine whether the order quantity is greater than the preset splitting threshold, and whether there are multiple available additive manufacturing machines capable of processing the product type corresponding to the order; S24: If the splitting conditions are met, dynamically determine the number N of devices to be allocated based on the order quantity and the size of the available device set; S25: Initialize the current load time for each of the selected N devices and calculate the dynamic batch size; S26: Under the condition that the remaining order quantity is greater than zero, iteratively and greedily allocate the remaining orders in a loop. Specifically, this includes: calculating the expected completion time after allocating the batch quantity to each device, selecting the device with the earliest expected completion time as the allocation device for the current round, allocating the batch quantity to the device, updating the device's load time and the remaining order quantity, until the order quantity is allocated. S27: Verify whether the difference in expected completion time of each sub-order for each device meets the preset balance constraint. If not, return to step S25 to adjust the batch size and redistribute until the constraint condition is met. S28: If a new order does not meet the splitting conditions, the new order will be directly added to the online order queue, waiting for subsequent equipment selection and scheduling execution.

[0009] Optionally, in step S22, the three-stage dynamic insertion scheduling strategy is as follows: New orders with extremely high priority are inserted at the front of the online order queue; High-priority new orders are inserted at the front of the online order queue; New orders with medium priority are inserted in the middle of the online order queue; New orders with low priority are inserted at the end of the online order queue.

[0010] Optionally, in step S26, the iterative greedy allocation is based on the principle of balanced expected completion time, minimizing the difference in expected completion time among the devices; the formula for calculating the expected completion time is: in, For orders In the equipment The expected completion time is as follows: For equipment Current load time, For orders In the equipment The unit product processing time, For orders Assigned to device The quantity.

[0011] Optionally, step S3 includes: S31: For orders to be dispatched, iterate through all additive manufacturing equipment that can process the product type corresponding to the order, build a set of candidate equipment, and obtain the current load time, historical usage count, and estimated processing time of each candidate equipment. S32: For each device in the candidate device set, calculate the corresponding individual score based on the comprehensive priority score, as well as the estimated completion time, processing time, and current load time; S33: Calculate the load balancing score based on the candidate device’s historical usage count, whether it is an unused device, and the relative relationship between the corresponding expected completion time and the current optimal expected completion time. Give positive rewards to unused devices and penalize devices with a high usage count. S34: The individual scores for maximum completion time, processing time, current load time, order priority, and long-term load balancing are weighted and summed to obtain the comprehensive score for each candidate device. The overall score is as follows: in, The overall score for device m. For the estimated completion time, For processing time, Scoring for load balancing The overall priority score for orders. This represents the historical number of times it has been used. , , , and These are the weighting coefficients corresponding to each objective; S35: Select the device with the highest overall score from all candidate devices as the target device; S36: Generate the corresponding production task on the target device, record the start time, end time, and processing quantity of the task, update the availability time and historical usage count of the target device, and remove the order from the online order queue.

[0012] Optionally, step S4 includes: The preset multi-level sorting rules include: The first level sorts orders by their processing time on the optimal candidate device from smallest to largest. The second level sorts processes from earliest to latest based on their expected completion time when the processing times are the same or the differences are less than a preset threshold. In the third level, when the processing time and the expected completion time are the same or the difference is less than the preset threshold, they are sorted from high to low according to the comprehensive priority score.

[0013] An electronic device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to enable the electronic device to perform an online scheduling method that considers load balancing for additive manufacturing.

[0014] A computer-readable storage medium storing instructions that, when executed, perform an online scheduling method for additive manufacturing that takes load balancing into account.

[0015] The beneficial effects of the technical solution provided in this application are: This invention addresses the online scheduling problem in additive manufacturing involving multiple factories, multiple devices, multiple product types, and variable batch sizes. Under conditions of dynamic order arrival, frequent production changes, and uncertain equipment status, the primary objective is to minimize the maximum completion time, while also considering comprehensive performance indicators such as equipment load balancing, response time for high-priority orders, and equipment resource utilization. By constructing a comprehensive priority model that integrates multi-dimensional features such as delivery date, order value, product importance, order size, waiting time, equipment load, and processing efficiency, and employing logarithmic compression normalization, features of different dimensions are uniformly mapped to a stable priority score, enabling fine-grained differentiation and differentiated scheduling of urgent and critical orders. Through a three-stage dynamic insertion strategy, orders of different priority levels are inserted into the front, middle, or rear of the online queue, avoiding the "long-term queuing of high-priority orders" problem caused by simple FIFO and single rules.

[0016] For large-volume orders, an iterative greedy algorithm for balancing expected completion time is introduced. Combined with dynamic batch size and parallel processing of multiple devices, the difference in expected completion time among devices is significantly reduced, fundamentally alleviating local bottlenecks and long-term load imbalance. In scenarios where orders are not split, a multi-objective comprehensive scoring method based on an index reward mechanism for unused devices is proposed. While ensuring overall completion time and processing efficiency, it prioritizes the use of long-term idle or rarely used devices, effectively improving global device utilization. The scheduling scheme optimized by the above methods significantly reduces the maximum completion time and improves the collaborative efficiency of multiple devices in actual simulations compared to traditional FIFO online scheduling. It can also intuitively display order flow and device load using Gantt charts, thus effectively solving the problems of untimely response, load imbalance, and insufficient resource utilization in existing additive manufacturing online scheduling. Attached Figure Description

[0017] The present application will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is a flowchart of the online scheduling method for additive manufacturing provided by the present invention; Figure 2 This is a flowchart illustrating the comprehensive priority calculation process of the online scheduling method for additive manufacturing provided by this invention. Figure 3 This invention provides an improved three-stage dynamic insertion flowchart for online order queues. Figure 4 This invention provides an improved multi-objective optimization device selection flowchart based on an unused device index reward mechanism. Figure 5 This is a Gantt chart of the first-come, first-served production plan of the online scheduling method for additive manufacturing provided by this invention; Figure 6 This invention provides an online scheduling method that solves the online scheduling problem for additive manufacturing. (Gantt chart) Figure 7 This is a schematic diagram of the electronic device structure in the embodiments of this application. Detailed Implementation

[0018] To provide a clearer understanding of the technical features, objectives, and effects of this application, the specific embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0019] Embodiments of this application provide an online scheduling method for additive manufacturing that considers load balancing.

[0020] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating the steps of an online scheduling method for additive manufacturing considering load balancing, as described in an embodiment of this application, including: S1: Obtain the order parameters of the new order and the current status parameters of the additive manufacturing equipment; construct a comprehensive priority model and calculate the comprehensive priority score P of the new order; S2: Based on the relationship between the comprehensive priority score P and the multi-level priority threshold, new orders are divided into different priority levels, and a segmented dynamic insertion strategy is used to insert them into different positions in the online order queue. S3: For orders to be dispatched, calculate the comprehensive score from the set of candidate equipment that can process the corresponding products, select the equipment with the best comprehensive score as the target equipment, and generate the production task on the target equipment. S4: Sort the orders to be scheduled in the online order queue according to the preset multi-level sorting rules, execute production scheduling in sequence, and update the equipment availability time, equipment historical usage count and the current system time until the orders to be scheduled at the current moment are processed.

[0021] As one example, this method is designed for distributed additive manufacturing environments with multiple factories, multiple devices, and multiple products. It is used to handle dynamic order scheduling problems in multi-variety, variable-batch, and large-scale additive manufacturing processes. It can respond to new orders in real time in highly dynamic and uncertain production environments, enabling rapid production changeover and optimization of scheduling schemes.

[0022] Step S1 includes: Order parameters must include at least one or more of the following: order arrival time, order quantity, basic priority, delivery date or delivery urgency, order value, and product importance; The equipment status parameters include at least one or more of the following: equipment availability time, equipment load, historical usage count, and processing efficiency; among which, equipment load is a status variable that characterizes the future occupancy level of the equipment. The comprehensive priority model is built based on multi-dimensional features, including static attribute features of orders and dynamic state features of the system; The overall priority score P is obtained by mapping the original overall score to the interval [0, 100] using a logarithmic compression normalization function.

[0023] Step S2 includes: S21: Based on the comprehensive priority score P and the preset priority threshold , , The relationship between them determines the priority level of new orders. ; among them when When is of extremely high priority, Time is of high priority, when When it is of medium priority, It is a low priority. S22: Based on priority levels, a three-stage dynamic insertion scheduling strategy is adopted to determine the insertion position of new orders in the online order queue; S23: Determine whether the order quantity is greater than the preset splitting threshold, and whether there are multiple available additive manufacturing machines capable of processing the product type corresponding to the order; S24: If the splitting conditions are met, dynamically determine the number N of devices to be allocated based on the order quantity and the size of the available device set; S25: Initialize the current load time for each of the selected N devices and calculate the dynamic batch size; S26: Under the condition that the remaining order quantity is greater than zero, iteratively and greedily allocate the remaining orders in a loop. Specifically, this includes: calculating the expected completion time after allocating the batch quantity to each device, selecting the device with the earliest expected completion time as the allocation device for the current round, allocating the batch quantity to the device, updating the device's load time and the remaining order quantity, until the order quantity is allocated. S27: Verify whether the difference in expected completion time of each sub-order for each device meets the preset balance constraint. If not, return to step S25 to adjust the batch size and redistribute until the constraint condition is met. S28: If a new order does not meet the splitting conditions, the new order will be directly added to the online order queue, waiting for subsequent equipment selection and scheduling execution.

[0024] In step S22, the three-stage dynamic insertion scheduling strategy is as follows: New orders with extremely high priority are inserted at the front of the online order queue; High-priority new orders are inserted at the front of the online order queue; New orders with medium priority are inserted in the middle of the online order queue; New orders with low priority are inserted at the end of the online order queue.

[0025] In step S26, the iterative greedy allocation is based on the principle of balanced expected completion time, minimizing the difference in expected completion time among the devices; the formula for calculating the expected completion time is: in, For orders In the equipment The expected completion time is as follows: For equipment Current load time, For orders In the equipment The unit product processing time, For orders Assigned to device The quantity.

[0026] Step S3 includes: S31: For orders to be dispatched, iterate through all additive manufacturing equipment that can process the product type corresponding to the order, build a set of candidate equipment, and obtain the current load time, historical usage count, and estimated processing time of each candidate equipment. S32: For each device in the candidate device set, calculate the corresponding individual score based on the comprehensive priority score, as well as the estimated completion time, processing time, and current load time; S33: Calculate the load balancing score based on the candidate device’s historical usage count, whether it is an unused device, and the relative relationship between the corresponding expected completion time and the current optimal expected completion time. Give positive rewards to unused devices and penalize devices with a high usage count. S34: The individual scores for maximum completion time, processing time, current load time, order priority, and long-term load balancing are weighted and summed to obtain the comprehensive score for each candidate device. The overall score is as follows: in, The overall score for device m. For the estimated completion time, For processing time, Scoring for load balancing The overall priority score for orders. This represents the historical number of times it has been used. , , , and These are the weighting coefficients corresponding to each objective; S35: Select the device with the highest overall score from all candidate devices as the target device; S36: Generate the corresponding production task on the target device, record the start time, end time, and processing quantity of the task, update the availability time and historical usage count of the target device, and remove the order from the online order queue.

[0027] Step S4 includes: The preset multi-level sorting rules include: The first level sorts orders by their processing time on the optimal candidate device from smallest to largest. The second level sorts processes from earliest to latest based on their expected completion time when the processing times are the same or the differences are less than a preset threshold. In the third level, when the processing time and the expected completion time are the same or the difference is less than the preset threshold, they are sorted from high to low according to the comprehensive priority score.

[0028] As one example, a set of scheduling schemes for the online scheduling problem of additive manufacturing were obtained through the online scheduling method, and a scheduling Gantt chart was drawn.

[0029] In some embodiments, the above-described online scheduling method for additive manufacturing can also be implemented in the following ways.

[0030] In this embodiment, the system needs to process n online orders, denoted as O={1,2,…,n}. Each order... Corresponding to a product type It includes the following attributes: arrival time Delivery period Order quantity Basic priority And the corresponding importance level of the product (imp_) The system is equipped with h additive manufacturing machines, and the equipment is grouped as follows: Each device It possesses the following characteristics: a set of processable product types. Unit product processing time Current available time A_{ }, and historical usage count U_{ Different machines may have varying processing efficiencies for the same product, and each machine already has some tasks in progress at the start of scheduling; therefore, their current availability and historical usage counts differ. The goal of scheduling is to allocate all orders to these additive manufacturing machines for production, maximizing the maximum completion time for all orders. Minimize. Simultaneously, the scheduling results must also consider comprehensive performance indicators such as equipment load balancing, response time of high-priority orders, and equipment resource utilization. Through comparison... Figure 5 and Figure 6The Gantt chart shown visually illustrates the difference in production efficiency between the First-Come, First-Served (FIFO) method and the online scheduling method proposed in this invention. This problem is based on a mixed-integer linear programming model and follows these assumptions: ① All additive manufacturing equipment is available from the initial moment, but existing tasks occupy a portion of the time; ② Once a production task begins processing, it cannot be interrupted or transferred to other equipment; ③ Each equipment can only process one production task at a time; ④ Orders can be split into multiple sub-tasks and assigned to different equipment for parallel processing; ⑤ The processing time of a production task is related to the number of orders, product type, and equipment type. The flowchart of this algorithm is shown below. Figure 1 As shown. The proposed rapid changeover online scheduling method for additive manufacturing specifically includes the following steps: Step 1: Establish a mixed-integer linear programming model for the rapid changeover online scheduling problem in additive manufacturing, with the main objective of minimizing the maximum completion time, while also taking into account equipment load balancing, response time of high-priority orders, and resource utilization; specifically, Step 1 includes Step 1.1, Step 1.2, and Step 1.3; Step 1.1: Define parameter symbols and decision variables. Parameter definitions include: i, Let j be the order number, n be the total number of orders, and the order set be O={1,2,…,n}; Let be the equipment number, h be the total number of equipment, and the equipment set be . Order-related parameters include: Let P represent the product type of order i, and let P be the set of product types. For equipment The set of product types that can be processed; Let i be the quantity of order i; Let i be the arrival time of order i; For order i, the delivery date is; The base priority for order i; imp_ For product type Importance level. Equipment and processing related parameters include: A_ For equipment The current available time; U_ For equipment Historical usage count; PT( , ) represents product type pi in the equipment Unit processing time; PT ( , , ) represents the allocation quantity for order i. In the equipment Upstream processed products Total processing time required; M is a maximum number. Policy and model parameters include: , , The priority threshold for the three-stage dynamic insertion strategy, and satisfies... ε is the threshold for the expected completion time balance constraint. , , , and For the weighting coefficients of the multi-objective comprehensive score, ; , , , , , , The weights of each feature item in the overall priority calculation are determined. Symbols related to the completion and time variables include: For the maximum completion time; Let i be the completion time of order i; , Order i in the device The start and completion times of the sub-tasks; For order i in the device The expected completion time; Let i be the overall priority score for order i; Let be the waiting time for order i. The decision variables are defined as follows: This is a binary variable, representing the subtask of order i on the device. Use 1 if processing is required, otherwise use 0; Let i be a non-negative integer variable, representing the order i assigned to the device. Quantity; Let i be a non-negative real number variable, representing order i in the device. The start time of the subtasks on the task; Let i be a non-negative real number variable, representing order i in the device. Subtask completion time; Let be a real number variable in the interval [0, 100], representing the overall priority score of order i; Let |Q| be an integer variable with values ​​of 0, 1, ..., |Q|, representing the insertion index of order i in the online order queue, where |Q| is the current queue length.

[0031] Step 1.2: Define the objective function to minimize the maximum completion time, while also considering auxiliary objectives. The main objective function is: The auxiliary objective function includes: minimizing the deviation in device usage frequency (load balancing). ,in For average usage frequency; minimize the expected completion time difference of split orders. ,in The number of devices allocated to order i. For order i, the set of available devices. The average expected completion time of order i across all devices; maximizing overall resource utilization. Minimize the waiting time for high-priority orders. Minimize penalties for late delivery ,in The penalty coefficient is constructed based on the urgency of delivery. ; Let be the completion time of order i.

[0032] Step 1.2: Construct the constraint set, including: Constraint 1: At any given time, each additive manufacturing equipment can only execute one production task, and the processing time intervals of any two production tasks assigned to the same equipment must not overlap on that equipment. Constraint 2: Each task can be assigned to at most one additive manufacturing device at any given time; Constraint 3: Orders can only be assigned to equipment capable of processing their corresponding product type; Constraint 4: Once a task begins, it must not be interrupted or moved to other equipment; processing must continue until completion. Constraint 5: The start time of a production task shall not be earlier than the available time of the allocated equipment, and its expression is: Constraint 6: The production task corresponding to an online order shall not start processing earlier than the order's arrival time; Constraint 7: The cumulative processing time of the equipment shall not exceed its maximum available processing time; Constraint 8: The completion time of a production task is equal to its start time plus the processing time on that equipment (which may include preparation time and post-processing time). Constraint 9: When an order is split into multiple sub-tasks, each sub-task must be assigned to a different additive manufacturing device; Constraint 10: All production tasks for an order must be completed before the order delivery date; Constraint 11: The number of subtasks that a single order can be split into does not exceed a preset limit, and the processing quantity of each subtask is not less than a preset minimum batch size; Constraint 12: To prevent low-priority orders from being starved for a long time, the overall priority of orders whose waiting time in the queue exceeds a preset threshold is dynamically increased to ensure that all orders can get processing opportunities within a limited time. Step 2: Based on the characteristics of the online scheduling problem for rapid changeover in additive manufacturing, an improved online heuristic scheduling algorithm is proposed; specifically, Step 2 includes Steps 2.1 to 2.8; Step 2.1: Input dataset information, including order set O, equipment set M, product type set P, static attributes such as arrival time, quantity, delivery date, basic priority, and product importance of each order, as well as status information such as the types of products that each equipment can process, current available time, and historical usage count; Step 2.2: Parameter initialization, set the weight parameters for comprehensive priority calculation, the threshold parameters for the three-stage dynamic insertion strategy, the large order splitting threshold (usually 30), the expected completion time balance constraint threshold ε (usually 10% of the maximum expected completion time), the weight coefficients for multi-objective comprehensive scoring, initialize the online order queue Q to be empty, and initialize the current system time t=0; Step 2.3: Establish an encoding scheme; the encoding scheme needs to solve three sub-problems: Problem 1: Order comprehensive priority calculation, mapping the static attributes and dynamic status of orders to a comprehensive priority score; Problem 2: Dynamic insertion into the online order queue, determining the position of an order in the queue based on the comprehensive priority score; Problem 3: Order splitting and device selection, splitting and allocating large orders to multiple devices, and selecting the optimal device for ordinary orders; In this embodiment, an event-driven online scheduling encoding scheme is designed, represented as η={ , ,…, },in The order number is used, and the coding order indicates the order arrival time or priority order; the coding scheme is as follows: Figure 2 As shown; Step 2.4: Initialize the system state; Specifically, this embodiment designs a heuristic strategy based on greedy algorithms and dynamic programming to initialize the initial in-process tasks, which includes: a heuristic strategy based on greedy algorithms and dynamic programming, including steps 2.4.1 to 2.4.4; Step 2.4.1: First, read the existing in-process tasks at the initial time of the system and initialize the available time of each device. Initialize the historical usage count for each device based on the completion time of the last task on each device. Step 2.4.2: For orders whose initial time has arrived but have not yet been dispatched, calculate the overall priority score for each order. The order is inserted into the online order queue Q according to the three-stage dynamic insertion strategy; Step 2.4.3: Sort the orders in the queue according to the processing time, expected completion time and comprehensive priority score in three levels; Step 2.4.4: Take the orders out of the queue in turn, perform equipment selection and task generation, and update the equipment status and system time; Second, random initialization: For new orders that arrive dynamically in the future, an online response mechanism is adopted to calculate the priority in real time and insert them into the queue; Step 2.5: Decode the encoding and calculate the fitness value of individuals in the initial population; specifically, for the decoding scheme, the process is as follows: Step 2.5.1: First, solve the problem of comprehensive priority calculation. When a new order arrives, obtain the static attributes of the order (basic priority, delivery date, product importance, order size) and the dynamic status of the system (order waiting time, current equipment load, processing efficiency), and calculate the static feature score. and dynamic feature scores The overall priority score is obtained by using a logarithmic compression normalization function. Step 2.5.2: Next, determine the queue insertion position based on the comprehensive priority score. With preset threshold , , Relationship, calculate queue position index Step 2.5.3: For large orders that meet the splitting conditions, an iterative greedy expected completion time balancing allocation algorithm is adopted to dynamically determine the number of devices N participating in the allocation, calculate the dynamic batch size, cyclically allocate batches to each device, update the expected completion time, until the order quantity is allocated, and verify whether the difference in expected completion time of each device meets the balancing constraint; Step 2.5.4: For orders that are not split or split subtasks, a multi-objective comprehensive score is calculated in the candidate device set, an unused device index reward mechanism is introduced, the device with the highest comprehensive score is selected as the target device, and a production task is generated; Step 2.6: Calculate the solutions according to their fitness values ​​(maximum completion time). Sort in descending order; the smaller the fitness value, the better the quality of the solution. Step 2.7: Order processing and equipment allocation are performed using an online scheduling event-driven approach. Specifically, the online scheduling operation of the algorithm is as follows: The online scheduling operator is an important operator in the rapid changeover online scheduling algorithm for additive manufacturing, simulating the dynamic order arrival and real-time response process. The online scheduling operation implements core functions such as order priority calculation, dynamic queue insertion, order splitting and allocation, and equipment selection. The online scheduling operator first needs to listen for new order arrival events. When a new order arrives, it triggers comprehensive priority calculation and queue insertion operations. Then, it determines whether the order meets the splitting conditions. If it does, it performs iterative greedy equalization of expected completion time; otherwise, it performs multi-target equipment selection. Finally, it generates production tasks and updates the system status. An event-driven mechanism is adopted in the coding scheduling, triggering corresponding scheduling decisions based on the order arrival time and the current system time. The scheduling operation process is as follows: Figure 3 As shown; Step 2.8: Sort the queue and assign tasks according to the multi-level sorting rules; specifically, the sorting and assignment operations of the algorithm are as follows: Sort the orders to be scheduled in the online order queue according to the preset multi-level sorting rules. The first level of sorting is based on the processing time of the order on the optimal candidate device. Sort by size from smallest to largest (SPT strategy), and then sort the second level according to the expected completion time if the processing time is the same or similar (difference <5%). Sort by time from earliest to latest. For third-level sorting, when processing time and estimated completion time are the same or similar, sort by overall priority score. Sort from highest to lowest; the sorting process is as follows: Figure 4 As shown; Step 2.9: Determine if there are any orders yet to be processed or any orders that will arrive in the future. If not, the algorithm terminates and outputs the scheduling result; otherwise, return to step 2.5 to continue processing newly arrived orders. Step 2.10: Output the scheduling results, including the production task allocation plan for all orders, the task sequence for each device, and the maximum completion time. Performance metrics such as equipment load balancing and high-priority order response time; algorithm flowcharts as follows Figure 1 As shown; Step 3: Through algorithm optimization, a set of scheduling schemes for the rapid production changeover online scheduling problem in additive manufacturing was obtained, and a scheduling Gantt chart was drawn; Specifically, step 3 includes: Step 3.1: After optimization, the algorithm obtains a complete scheduling scheme, including the production task allocation of all orders on each device, the start and end times of the tasks, the task sequence of each device, and system performance indicators; Step 3.2: A Gantt chart is drawn based on the generated scheduling scheme. The horizontal axis of the Gantt chart represents time, and the vertical axis represents devices. Rectangles of different colors represent the processing intervals of different orders or sub-tasks on each device, and key information such as maximum completion time and device utilization rate are marked on the chart; Step 3.3: The Gantt chart of the online scheduling method of this invention is compared with the Gantt chart of the FIFO method to demonstrate the advantages of this invention in terms of maximum completion time, device load balancing, and response time of high-priority orders. The comparison results are as follows: Figure 5 and Figure 6 As shown.

[0033] This application also discloses an electronic device. (See reference...) Figure 7 , Figure 7 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application. The electronic device 500 may include: at least one processor 501, at least one network interface 504, a user interface 503, a memory 505, and at least one communication bus 502.

[0034] The communication bus 502 is used to enable communication between these components.

[0035] The user interface 503 may include a display screen, and optionally, the user interface 503 may also include a standard wired interface or a wireless interface.

[0036] The network interface 504 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0037] This application also discloses a computer-readable storage medium storing a plurality of instructions adapted for loading by a processor to execute the above-described online scheduling method for additive manufacturing with load balancing considerations.

[0038] The above are merely exemplary embodiments of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure.

[0039] This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. An online scheduling method for additive manufacturing that considers load balancing, characterized in that, The method includes the following steps: S1: Obtain the order parameters of the new order and the current status parameters of the additive manufacturing equipment; construct a comprehensive priority model and calculate the comprehensive priority score P of the new order; S2: Based on the relationship between the comprehensive priority score P and the multi-level priority threshold, new orders are divided into different priority levels, and a segmented dynamic insertion strategy is used to insert them into different positions in the online order queue. S3: For orders to be dispatched, calculate the comprehensive score from the set of candidate equipment that can process the corresponding products, select the equipment with the best comprehensive score as the target equipment, and generate the production task on the target equipment. S4: Sort the orders to be scheduled in the online order queue according to the preset multi-level sorting rules, execute production scheduling in sequence, and update the equipment availability time, equipment historical usage count and the current system time until the orders to be scheduled at the current moment are processed.

2. The online scheduling method for additive manufacturing considering load balancing as described in claim 1, characterized in that, Step S1 includes: Order parameters must include at least one or more of the following: order arrival time, order quantity, basic priority, delivery date or delivery urgency, order value, and product importance; The equipment status parameters include at least one or more of the following: equipment availability time, equipment load, historical usage count, and processing efficiency; among which, equipment load is a status variable that characterizes the future occupancy level of the equipment. The comprehensive priority model is built based on multi-dimensional features, including static attribute features of orders and dynamic state features of the system; The overall priority score P is obtained by mapping the original overall score to the interval [0, 100] using a logarithmic compression normalization function.

3. The online scheduling method for additive manufacturing considering load balancing as described in claim 2, characterized in that, Step S2 includes: S21: Based on the comprehensive priority score P and the preset priority threshold , , The relationship between them determines the priority level of new orders. ; among them when When is of extremely high priority, Time is of high priority, when When it is of medium priority, It is a low priority. S22: Based on priority levels, a three-stage dynamic insertion scheduling strategy is adopted to determine the insertion position of new orders in the online order queue; S23: Determine whether the order quantity is greater than the preset splitting threshold, and whether there are multiple available additive manufacturing machines capable of processing the product type corresponding to the order; S24: If the splitting conditions are met, dynamically determine the number N of devices to be allocated based on the order quantity and the size of the available device set; S25: Initialize the current load time for each of the selected N devices and calculate the dynamic batch size; S26: Under the condition that the remaining order quantity is greater than zero, iteratively and greedily allocate the remaining orders in a loop. Specifically, this includes: calculating the expected completion time after allocating the batch quantity to each device, selecting the device with the earliest expected completion time as the allocation device for the current round, allocating the batch quantity to the device, updating the device's load time and the remaining order quantity, until the order quantity is allocated. S27: Verify whether the difference in expected completion time of each sub-order for each device meets the preset balance constraint. If not, return to step S25 to adjust the batch size and redistribute until the constraint condition is met. S28: If a new order does not meet the splitting conditions, the new order will be directly added to the online order queue, waiting for subsequent equipment selection and scheduling execution.

4. An online scheduling method for additive manufacturing considering load balancing as described in claim 3, characterized in that, In step S22, the three-stage dynamic insertion scheduling strategy is as follows: New orders with extremely high priority are inserted at the front of the online order queue; High-priority new orders are inserted at the front of the online order queue; New orders with medium priority are inserted in the middle of the online order queue; New orders with low priority are inserted at the end of the online order queue.

5. An online scheduling method for additive manufacturing considering load balancing as described in claim 1, characterized in that, In step S26, the iterative greedy allocation is based on the principle of balanced expected completion time, minimizing the difference in expected completion time among the devices; the formula for calculating the expected completion time is: in, For orders In the equipment The expected completion time is as follows: For equipment Current load time, For orders In the equipment The unit product processing time, For orders Assigned to device The quantity.

6. An online scheduling method for additive manufacturing considering load balancing as described in claim 1, characterized in that, Step S3 includes: S31: For orders to be dispatched, iterate through all additive manufacturing equipment that can process the product type corresponding to the order, build a set of candidate equipment, and obtain the current load time, historical usage count, and estimated processing time of each candidate equipment. S32: For each device in the candidate device set, calculate the corresponding individual score based on the comprehensive priority score, as well as the estimated completion time, processing time, and current load time; S33: Calculate the load balancing score based on the candidate device’s historical usage count, whether it is an unused device, and the relative relationship between the corresponding expected completion time and the current optimal expected completion time. Give positive rewards to unused devices and penalize devices with a high usage count. S34: The individual scores for maximum completion time, processing time, current load time, order priority, and long-term load balancing are weighted and summed to obtain the comprehensive score for each candidate device. The overall score is as follows: in, The overall score for device m. For the estimated completion time, For processing time, Scoring for load balancing The overall priority score for orders. This represents the historical number of times it has been used. , , , and These are the weighting coefficients corresponding to each objective; S35: Select the device with the highest overall score from all candidate devices as the target device; S36: Generate the corresponding production task on the target device, record the start time, end time, and processing quantity of the task, update the availability time and historical usage count of the target device, and remove the order from the online order queue.

7. An online scheduling method for additive manufacturing considering load balancing as described in claim 1, characterized in that, Step S4 includes: The preset multi-level sorting rules include: The first level sorts orders by their processing time on the optimal candidate device from smallest to largest. The second level sorts processes from earliest to latest based on their expected completion time when the processing times are the same or the differences are less than a preset threshold. In the third level, when the processing time and the expected completion time are the same or the difference is less than the preset threshold, they are sorted from high to low according to the comprehensive priority score.

8. An electronic device, characterized in that, It includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device performs the online scheduling method for additive manufacturing that takes load balancing into account, as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed by a computer, perform an online scheduling method for additive manufacturing that takes load balancing into account, as described in any one of claims 1-7.

Images