Intelligent production scheduling method and system for cable flexible manufacturing

Abstract

The invention discloses an intelligent production scheduling method and system for cable flexible manufacturing, and relates to the field of production management scheduling, and the method comprises the steps: firstly, through data solidification, comprehensively capturing dynamic information such as an emergency order, a current plan, a factory state and model changing time, and forming a context basis of a decision; then, a rapid heuristic algorithm is used to carry out cost quantitative evaluation on all possible insertion schemes of the emergency order, and a high-quality initial scheduling solution is rapidly generated; and finally, introducing a simulated annealing element heuristic algorithm, and carrying out global iterative search by taking the initial solution as a starting point so as to find an optimal solution with minimum disturbance to the whole production plan. According to the method, a complex experience decision is converted into a data-driven automatic optimization process, rapid and accurate adjustment of the production plan is realized, and the overall stability and efficiency of the production line are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of production management scheduling, and more specifically, to an intelligent production scheduling method and system for cable flexible manufacturing. BACKGROUND

[0002] With the intensification of market competition and the development of personalized customer demand, cable manufacturing industry is transforming towards a flexible production mode of multi-variety, small batch and fast delivery. In this mode, the stability of production planning faces great challenges, especially the frequent occurrence of high-priority emergency orders, which causes serious impact on the original production order. How to quickly respond and handle these emergency orders while ensuring the efficiency and stability of the entire production system has become the key to improving the core competitiveness of cable enterprises.

[0003] Currently, cable manufacturing enterprises mainly rely on the personal experience of production schedulers to manually adjust or use some automatic scheduling systems based on simple rules when dealing with such emergency orders. The manual scheduling method is highly subjective and slow in response, and it is difficult to comprehensively evaluate the cascading effects of the inserted order on other order delivery dates, equipment changeover costs and production rhythm in a short time, often leading to trade-offs and causing more widespread production delays and cost increases. The existing automatic scheduling methods, such as traditional local search algorithms, are not efficient in dealing with cable manufacturing problems with specific constraints and cost structures. These methods usually use general neighborhood operations (such as random insertion, exchange), lack deep adaptation to cable production-specific processes (such as changeover time matrix), cannot accurately quantify the comprehensive impact of the inserted order decision, are prone to local optimal solutions, and are difficult to find a truly balanced scheduling solution for all parties.

[0004] Therefore, how to intelligently insert a high-priority emergency order into the planned production sequence without seriously damaging the existing production rhythm and meeting the delivery date, and quantify its impact on other orders, i.e., the dynamic embedding and impact minimization problem of emergency insertion, has become a technical problem to be solved in the current cable flexible manufacturing field. The existing technology fails to provide an intelligent scheduling method that can systematically evaluate the impact of the inserted order and perform global optimization search, resulting in the handling of emergency orders being a pain point in production management. SUMMARY

[0005] To solve the problems in the background art, according to an aspect of the present application, an intelligent production scheduling method for cable flexible manufacturing is provided, which comprises:

[0006] obtaining emergency order data, current production plan, real-time factory state and changeover time matrix;

[0007] data solidification on the emergency order data, the current production plan, the real-time factory status and the changeover time matrix to obtain a problem context;

[0008] fast heuristics initialization on the emergency order data, the current production plan based on the problem context to obtain an initial solution;

[0009] simulated annealing based meta-heuristics iterative search on the initial solution based on simulated annealing parameters to obtain an optimal solution, the simulated annealing parameters including an initial temperature, a termination temperature and a cooling rate;

[0010] conversion of the optimal solution into a new set of scheduling instructions executable by the MES system.

[0011] According to another aspect of the present application, there is provided an intelligent production scheduling system for cable flexible manufacturing, comprising:

[0012] a production data acquisition module configured to acquire emergency order data, a current production plan, real-time factory status and a changeover time matrix;

[0013] a data solidification module configured to perform data solidification on the emergency order data, the current production plan, the real-time factory status and the changeover time matrix to obtain a problem context;

[0014] an initial solution generation module configured to perform fast heuristics initialization on the emergency order data, the current production plan based on the problem context to obtain an initial solution;

[0015] an optimal solution generation module configured to perform simulated annealing based meta-heuristics iterative search on the initial solution based on simulated annealing parameters to obtain an optimal solution, the simulated annealing parameters including an initial temperature, a termination temperature and a cooling rate;

[0016] a scheduling module configured to convert the optimal solution into a new set of scheduling instructions executable by the MES system.

[0017] Compared with the prior art, the intelligent production scheduling method and system for cable flexible manufacturing provided by the application solve the problem of dynamic embedding and influence minimization of emergency insertion in cable flexible manufacturing, and adopt a two-stage strategy of fast initialization + global optimization. First, through data solidification, dynamic information such as emergency orders, current plans, factory states and changeover times is comprehensively captured to form the context basis for decision-making. Then, a fast heuristic algorithm is used to quantitatively evaluate the cost of all possible insertion schemes of the emergency order, and a high-quality initial scheduling solution is quickly generated to avoid blind search. Finally, a simulated annealing metaheuristic algorithm is introduced to perform global iterative search starting from the initial solution, and the probabilistic jump-out mechanism can effectively avoid the local optimal trap that the traditional method is prone to fall into, so as to find the optimal solution with the minimum disturbance to the overall production plan. The method changes the complex experience decision-making into a data-driven automated optimization process, realizes the rapid and accurate adjustment of the production plan, and improves the overall stability and efficiency of the production line. BRIEF DESCRIPTION OF DRAWINGS

[0018] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description of embodiments of the present application taken in conjunction with the accompanying drawings.

[0019] Figure 1 A flowchart of the intelligent production scheduling method for cable flexible manufacturing according to the embodiments of the present application.

[0020] Figure 2 A data flow diagram of the intelligent production scheduling method for cable flexible manufacturing according to the embodiments of the present application.

[0021] Figure 3 A flowchart of step 3 in the intelligent production scheduling method for cable flexible manufacturing according to the embodiments of the present application.

[0022] Figure 4 A flowchart of step 4 in the intelligent production scheduling method for cable flexible manufacturing according to the embodiments of the present application.

[0023] Figure 5 A block diagram of the intelligent production scheduling system for cable flexible manufacturing according to the embodiments of the present application. DETAILED DESCRIPTION

[0024] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. It should be understood that the accompanying drawings and embodiments of the present disclosure are only for exemplary purposes, and are not intended to limit the protection scope of the present disclosure.

[0025] Based on the deficiencies in the above technical field, the present application provides an intelligent production scheduling method for cable flexible manufacturing. Figure 1A flow chart of the intelligent production scheduling method for cable flexible manufacturing according to an embodiment of the present application. Figure 2 A data flow diagram of the intelligent production scheduling method for cable flexible manufacturing according to an embodiment of the present application. As shown in Figure 1 and Figure 2 The intelligent production scheduling method for cable flexible manufacturing according to an embodiment of the present application includes: step 1, obtaining emergency order data, current production plan, real-time factory status and changeover time matrix; step 2, data solidification of the emergency order data, current production plan, real-time factory status and changeover time matrix to obtain problem context; step 3, based on the problem context, quick heuristic initialization of the emergency order data and current production plan to obtain an initial solution; step 4, meta-heuristic iterative search based on simulated annealing of the initial solution to obtain an optimal solution, the simulated annealing parameters including initial temperature, termination temperature and cooling rate; and step 5, converting the optimal solution into a new set of scheduling instructions executable by the MES system.

[0026] In step 1, the emergency order data, current production plan, real-time factory status and changeover time matrix are obtained. It should be understood that in the cable flexible manufacturing environment of multi-variety and small batch, the production plan is frequently impacted by high-priority emergency orders. In order to intelligently embed new orders into the existing plan and minimize the negative impact on the overall production order, a comprehensive, accurate and static decision basis is needed. Any scheduling calculation that deviates from the real-time production state may lead to incorrect judgments and exacerbate production delays and cost waste. Therefore, before starting any complex optimization algorithm, the first task is to accurately capture all relevant information at the decision moment and solidify the dynamic production environment into a static, analyzable data snapshot. Therefore, in order to build such a complete and reliable problem context to provide a solid data cornerstone for subsequent heuristic initialization and global optimization search, the present application first obtains the emergency order data, current production plan, real-time factory status and changeover time matrix.

[0027] In one possible implementation, step 1 operates as follows: To clearly describe the data acquisition process, a specific scenario is set: at 10:00 am on September 26, 2025, a high-priority emergency order request is triggered. At this time, the intelligent production scheduling method for cable flexible manufacturing begins to execute the first step, data acquisition and solidification. First, interact with the order management module to obtain detailed data of the emergency order. The emergency order data is a structured information body, containing the unique identification of the order, the product specification required for production, the order quantity, and the required delivery deadline. For example, the emergency order data obtained this time is: {order number: "JJ-20250926-001", product specification: "YJV-1kV-4x185", quantity: "500 meters", delivery date: "September 28, 2025"}. This set of data clearly defines the core goal of this scheduling optimization.

[0028] Next, in order to understand the baseline state of the current production, it is necessary to retrieve the current production plan that has been planned from the manufacturing execution module (MES). The current production plan is an ordered sequence containing multiple orders, and details the production arrangement and estimated completion time of each order. This plan is the basis for evaluating the impact of inserting orders. For example, the plan for production line L1 obtained from the manufacturing execution module is: order one (order number: "regular order A", product specification: "YJV-1kV-4x95", planned completion time: "September 26, 2025 18:00"), followed by order two (order number: "regular order B", product specification: "YJV-1kV-4x120", planned completion time: "September 27, 2025 12:00") and order three (order number: "regular order C", product specification: "YJV-1kV-4x240", planned completion time: "September 28, 2025 10:00").

[0029] While obtaining the macro plan, it is also necessary to capture the micro-factory state at the time the request is initiated, i.e. 10:00 am, which is crucial for accurate calculation. Real-time factory state is a momentary freeze of the physical state of the production site, including equipment operation and work-in-process progress. This information is obtained by querying the production line's data acquisition and monitoring control component (SCADA) or device PLC. For example, at 10:00 am, the real-time factory state obtained is: production line L1 is executing regular order A, with 30% of the total task completed, and the equipment is running normally. The locking of this state means that all subsequent calculations will be based on the accurate fact that regular order A has 70% left to complete, ensuring the timeliness of the decision.

[0030] Finally, to quantify the cost of switching between different production tasks, the changeover time matrix needs to be obtained from the process database. The changeover time matrix is a pre-defined two-dimensional table that defines the preparation time required to switch from producing one product specification to another based on historical data and process standards. It is a key input for calculating the total changeover cost. The matrix is relatively static data, but it has a huge impact on scheduling decisions. For example, the obtained matrix fragment can show that switching from YJV-1kV-4x95 regular order A to YJV-1kV-4x185 emergency order requires 1.5 hours; while switching from YJV-1kV-4x185 to YJV-1kV-4x120 regular order B requires 1.2 hours.

[0031] In step 2, the emergency order data, the current production plan, the real-time factory status, and the changeover time matrix are data solidified to obtain the problem context. Accordingly, in the intelligent scheduling decision-making process, the real-time and consistency of the input data are the premise to ensure the effectiveness of the results. The production site is a dynamic environment, and information such as equipment status and work-in-process progress is updated every moment. If the underlying data is still changing while performing complex scheduling calculations, the calculation process will lose its stable benchmark, and the scheduling scheme obtained may be outdated or not feasible as soon as it is generated. Therefore, a static and consistent data set needs to be formed at the moment when the optimization algorithm is started. To this end, the application performs data solidification on the emergency order data, the current production plan, the real-time factory status, and the changeover time matrix to obtain the problem context, which can build such a stable and reliable calculation snapshot, ensuring that all subsequent decisions are based on accurate information from the same time section.

[0032] In a feasible solution, step 2 operates as follows: first, the emergency order data obtained in the previous step {order number: "JJ-20250926-001", product specification: "YJV-1kV-4x185", quantity: "500 meters", delivery date: "September 28, 2025"} is directly placed into the problem context.

[0033] Next, the current production plan and real-time factory status are fused. It is known that at 10:00 am, "regular order A" has completed 30%, which means that the original plan needs to be corrected to reflect this real-time progress. Therefore, the remaining production time of regular order A is recalculated, and its estimated completion time is updated. After that, this corrected production plan, which is an ordered sequence consisting of 70% of regular order A, regular order B, and regular order C, is taken as the baseline plan in the problem context. At the same time, the obtained changeover time matrix is also fully incorporated into the problem context.

[0034] Finally, a problem context data object containing all necessary information is generated. This object has a clear internal structure, including key fields such as emergency order details, baseline plan sequences, and changeover time costs. For example, the output problem context is a comprehensive data package containing the aforementioned emergency order information, the adjusted sequences of three regular orders and their estimated completion times, and the time matrix required for switching between product specifications.

[0035] In step 3, based on the problem context, a rapid heuristic initialization is performed on the emergency order data and the current production plan to obtain an initial solution. It is understandable that, faced with the complex scheduling problem caused by emergency order insertion, directly using a global search algorithm (such as simulated annealing) to start from a random or arbitrary initial state would result in a lengthy and inefficient search process, easily getting trapped in local optima. A high-quality initial solution can provide a superior starting point for the global optimization algorithm, significantly improving convergence speed and solution quality. Therefore, before initiating a complex iterative search, an efficient and fast method is needed to initially screen and evaluate all possible insertion schemes, quickly identifying a candidate scheme with high potential. To this end, this application uses a rapid heuristic initialization of the emergency order data and the current production plan based on the problem context to quickly generate a relatively low-cost, competitive initial solution from the vast solution space, thus laying a solid foundation for the subsequent metaheuristic iterative search and ensuring that the global optimum is ultimately found efficiently.

[0036] In one feasible solution, Figure 3 This is a flowchart of step 3 in the intelligent production scheduling method for flexible cable manufacturing according to an embodiment of this application. Figure 3 As shown, step 3, based on the problem context, performs rapid heuristic initialization on the emergency order data and the current production plan to obtain an initial solution, including: step 31, generating n+1 logical insertion points based on the length n of the current production plan; step 32, inserting the emergency order data into each of the logical insertion points to obtain a candidate plan sequence list; step 33, based on the problem context, quantifying the impact cost of each candidate plan in the candidate plan sequence list to obtain a scored candidate plan list; step 34, performing a minimum cost search on the scored candidate plan list to obtain an initial solution.

[0037] In the above scheme, step 3 is operated as follows: first, start step 31. In the problem context, the known baseline schedule sequence is regular order A remaining 70%, regular order B, and regular order C, so the length n of this schedule is 3. Accordingly, n+1, i.e. 4, logical insertion points will be generated. These 4 insertion points are not physical positions, but represent all possible before-and-after order relationships of the urgent order in the existing order sequence. The first insertion point is before regular order A remaining 70%, representing that the urgent order becomes the current most prioritized task; the second insertion point is between regular order A remaining 70% and regular order B; the third insertion point is between regular order B and regular order C; and the fourth insertion point is after regular order C, representing that the urgent order is executed after all planned orders are completed. These 4 logical insertion points exhaust all possibilities of the insertion of the urgent order.

[0038] Subsequently, step 32 is entered. The urgent order “JJ-20250926-001” contained in the problem context is placed into each logical insertion point in turn, thereby forming 4 completely new, temporary production schedule sequences. Inserting the urgent order into the first point generates candidate schedule one: “urgent order JJ-20250926-001” -> “regular order A remaining 70%” -> “regular order B” -> “regular order C”. Inserting it into the second point generates candidate schedule two: “regular order A remaining 70%” -> “urgent order JJ-20250926-001” -> “regular order B” -> “regular order C”. And so on, finally generating candidate schedule three: “regular order A remaining 70%” -> “regular order B” -> “urgent order JJ-20250926-001” -> “regular order C”, and candidate schedule four: “regular order A remaining 70%” -> “regular order B” -> “regular order C” -> “urgent order JJ-20250926-001”. These four candidate schedules collectively constitute the output candidate schedule sequence list, which fully presents all preliminary scheduling schemes.

[0039] To transform the abstract scheduling sequence into a concrete cost score, thereby providing a clear and quantifiable decision-making basis for the subsequent minimum cost search, and ensuring that the selected initial solution is a result that comprehensively considers delivery time, efficiency, and stability, proceed to step 33. In a feasible solution, the impact cost of each candidate plan includes weighted delay penalty, total changeover cost, and plan disturbance penalty. To illustrate the calculation process in detail, candidate plan two, namely the sequence "Regular Order A (70% remaining)" -> "Urgent Order JJ-20250926-001" -> "Regular Order B" -> "Regular Order C", is used as an example for calculation. First, calculate the total changeover cost. Total changeover cost is a key indicator for measuring the continuity of the production process; it represents the total downtime and adjustment time caused by producing different specifications of products. In a feasible solution, the total changeover cost is calculated using the following formula:

[0040]

[0041] in, For the transformation time function, For the first Current orders for each candidate program, For the first Among the candidate plans Pre-orders, For candidate plans The total replacement cost. This formula means that for a given planning sequence... Its total changeover cost is equal to the changeover time between all adjacent orders in the sequence. The sum of the costs. In this example, three changeover times need to be calculated: the time to switch from product specification "YJV-1kV-4x95" in "Regular Order A" to "YJV-1kV-4x185" in "Urgent Order"; the time to switch from "Urgent Order" to "Regular Order B" to "YJV-1kV-4x120"; and the time to switch from "Regular Order B" to "Regular Order C" to "YJV-1kV-4x240". By querying the changeover time matrix in the problem context, the times required for these three changes are obtained, such as 1.5 hours, 1.2 hours, and 0.8 hours respectively. Therefore, the total changeover cost of Candidate Plan Two is... The sum of these three is 1.5 + 1.2 + 0.8 = 3.5 hours.

[0042] Next, the weighted delay penalty is calculated. The weighted delay penalty quantifies the business loss caused by order delays due to plan changes; it considers not only the duration of the delay but also the importance of different orders. In one feasible approach, the weighted delay penalty is calculated using the following formula:

[0043]

[0044] in, The first in the candidate plan sequence list One candidate program, For the first The first of the candidate plans One order, For the first Order completion time for each order For the first Order priority for each order. For the first Delivery time for each order To find the maximum value function, For candidate plans The weighted delay penalty. This formula shows that the total penalty value is the sum of the individual penalty values ​​of all orders in the sequence. The penalty value of each order is equal to its priority. Multiply by its actual completion time Beyond its delivery date Duration. Function This ensures that only orders that are actually overdue are penalized. Before calculation, each order needs to be prioritized; this priority can be determined based on customer level or contractual agreement. For example, urgent orders have a priority of 10, while regular orders A, B, and C have priorities of 3, 2, and 1, respectively. Then, based on the sequence order, production time of each order, and changeover time, the new completion time for each order under Candidate Plan Two needs to be calculated. Calculations show that both urgent order and regular order A were on time, but the new completion time for regular order B was 6 hours later than its delivery date, and for regular order C it was 8 hours later. Therefore, the weighted delay penalty for candidate plan two... =(10*0)+(3*0)+(2*6)+(1*8)=20.

[0045] Next, the plan disturbance penalty is calculated. The plan disturbance penalty measures the impact of order insertions on the stability of the original production plan, reflecting the management costs of plan changes and the disruption to production takt time. In one feasible approach, the plan disturbance penalty is calculated using the following formula:

[0046]

[0047] in, For the number of orders affected, This is the average delay time. For candidate plans The planned disturbance penalty. This formula means that the disturbance penalty equals the number of orders affected. Multiply by the average delay time of these orders The affected orders are those whose estimated completion times are later than their original baseline plan after the insertion of the urgent order. In this example, the production of regular order A is in progress and its completion time is not affected, while the completion times of regular orders B and C are both delayed, so the number of affected orders is 2. Their delay durations are 6 hours and 8 hours respectively, and the average delay time is (6+8) / 2 = 7 hours. Therefore, the plan disturbance penalty of candidate plan two is 2*7 = 14.

[0048] Finally, the three cost components are added together to get the total impact cost of candidate plan two as 3.5+20+14 = 37.5. Similarly, the exact same quantification process is performed on the remaining three candidate plans (candidate plan one, three, and four) in the list, and their total impact costs are calculated respectively. The final output is a scored candidate plan list, where each candidate plan is associated with a clear cost value, for example: [{candidate plan one, cost: 55.0}, {candidate plan two, cost: 37.5}, {candidate plan three, cost: 42.0}, {candidate plan four, cost: 68.0}].

[0049] Finally, step 34 is performed. This search process is a direct comparison and selection operation. Each element in the list is traversed. First, the first element candidate plan one and its cost 55.0 are tentatively set as the current optimal solution. Then, the second element candidate plan two is processed, which has a cost of 37.5. Since 37.5 is less than 55.0, the current optimal solution is updated to candidate plan two. Then, the third element candidate plan three is processed, which has a cost of 42.0, which is greater than the currently recorded minimum cost 37.5, so no update is made. Finally, the fourth element candidate plan four is processed, which has a cost of 68.0, also greater than 37.5, so no update is made. After traversing all candidate plans, the search ends. The candidate plan two is determined as the solution with the lowest cost. Therefore, the output initial solution is a specific scheduling scheme, i.e., the production sequence: “regular order A remaining 70%” -> “urgent order JJ-20250926-001” -> “regular order B” -> “regular order C”, with an associated comprehensive impact cost of 37.5. This initial solution is not the final scheme, but it is the most promising solution with the smallest impact on the existing plan after the fast heuristic evaluation, and will be used as the starting point for the simulated annealing meta-heuristic iterative search in the next stage.

[0050] ​​​In step 4, based on the simulated annealing parameters, a metaheuristic iterative search based on simulated annealing is performed on the initial solution to obtain the optimal solution. The simulated annealing parameters include the initial temperature, the termination temperature, and the cooling rate. It should be understood that the initial solution obtained through rapid heuristic initialization, although having the lowest cost among all direct insertion schemes, is essentially a local optimum and does not explore broader scheduling possibilities, such as creating more favorable insertion conditions for urgent orders by adjusting the order of existing orders. Cable production scheduling is a complex combinatorial optimization problem with a huge solution space. Simply using insertion search may miss the globally optimal solution. To overcome the limitation of local optima, an algorithm capable of global search is needed, which should have the ability to accept poor solutions with a certain probability, thereby escaping local traps and exploring a broader solution space. Therefore, to perform deep optimization based on the initial solution, this application performs a metaheuristic iterative search on the initial solution based on simulated annealing parameters. That is, by simulating the cooling and state stabilization process during physical annealing, it searches for the optimal scheduling scheme that minimizes the impact on the production plan globally in a probabilistic optimization manner.

[0051] In one feasible solution, Figure 4 This is a flowchart of step 4 in the intelligent production scheduling method for flexible cable manufacturing according to an embodiment of this application. Figure 4 As shown, step 4 involves performing a metaheuristic iterative search based on simulated annealing parameters on the initial solution to obtain the optimal solution. The simulated annealing parameters include the initial temperature, the termination temperature, and the cooling rate. This includes: step 41, initializing the search state based on the simulated annealing parameters and the initial solution to obtain the current solution state, the optimal solution state, and the current temperature; step 42, performing a single-iteration search and state update on the current solution state, the optimal solution state, and the current temperature based on the neighborhood generation function and the influence cost function to obtain the updated current solution state and the updated optimal solution state; and step 43, performing temperature annealing and cyclic control on the updated current solution state and the updated optimal solution state to obtain the optimal solution.

[0052] In the above scheme, step 4 is operated as follows: Firstly, start step 41. The simulated annealing parameters are the key values that control the behavior of the whole search process, including the initial temperature, the final temperature and the cooling rate. The initial temperature is a higher value, which is used to allow the algorithm to accept non-optimal moves with a higher probability in the early stage of the search, so as to carry out extensive exploration; the final temperature is a very small value close to zero, which is used as the condition for the algorithm to stop; the cooling rate is a coefficient between 0 and 1, which determines the speed of temperature drop. These parameters are determined based on experience or through experimental debugging of specific problems. In this example, the initial temperature is set to 100, the final temperature is set to 0.1, and the cooling rate is set to 0.95. The initialization process first needs to determine the cost of the initial solution. By calling the impact cost function used in step 33, or directly from the output of the previous step, it is confirmed that the impact cost of the initial solution is 37.5. Next, start assigning values to the search state. Firstly, the current solution state, which represents the current position of the algorithm in the search space, is set. The sequence in the current solution state, denoted as S_current, is set to the sequence of the initial solution, i.e. "regular order A remaining 70%" -> "emergency order JJ-20250926-001" -> "regular order B" -> "regular order C", and its cost, denoted as Cost_current, is set to 37.5. Then, the optimal solution state, which is used to record the solution with the lowest cost found during the whole search process, is set. In the initialization stage, no search has been performed, so the best solution found is the initial solution itself. Therefore, the sequence in the optimal solution state, denoted as S_best, and the cost, denoted as Cost_best, are also set to the sequence and cost of the initial solution, i.e. 37.5. Finally, the temperature is initialized. The initial temperature value is read from the preset simulated annealing parameters and assigned to the current temperature variable. Therefore, the current temperature is set to 100. After the above operations, the output is three initialized state variables: the current solution state {sequence: "regular order A..." -> "emergency order...", cost: 37.5}, the optimal solution state {sequence: "regular order A..." -> "emergency order...", cost: 37.5}, and the current temperature 100. These three variables provide a complete starting point for the subsequent single iteration search and state update.

[0053] After that, step 42 is entered to gradually converge from the initial solution to the global optimal solution. In a feasible solution, step 42 performs a single iteration search and state update on the current solution state, the optimal solution state and the current temperature based on the neighborhood generating function and the impact cost function to obtain the updated current solution state and the updated optimal solution state, including: step 421, inputting the current solution state into the neighborhood generating function to obtain a new neighborhood solution; step 422, inputting the new neighborhood solution into the impact cost function to obtain the impact cost of the neighborhood solution; step 423, calculating the cost difference between the impact cost of the neighborhood solution and the impact cost of the current solution; step 424, if the cost difference is less than 0, directly updating the current solution state to the combination of the new neighborhood solution and the impact cost of the neighborhood solution; if the cost difference is greater than or equal to zero, calculating the probability of accepting the neighborhood solution and determining whether to update the current solution state to the combination of the new neighborhood solution and the impact cost of the neighborhood solution; step 425, comparing the impact cost of the neighborhood solution with the impact cost of the optimal solution, if the impact cost of the neighborhood solution is less than the impact cost of the optimal solution, updating the optimal solution state to the combination of the new neighborhood solution and the impact cost of the neighborhood solution.

[0054] Specifically, the process first enters step 421, which is completed by the neighborhood generating function. The neighborhood generating function is not a single operation, but a composite function containing multiple different perturbation strategies (called neighborhood operators), which can dynamically select an operator to transform the current solution according to the preset rules. The execution of the neighborhood generating function is divided into two stages: neighborhood operator dynamic selection and neighborhood solution sequence construction. First, enter the neighborhood operator dynamic selection stage. The goal of this stage is to decide which way to change the current order sequence. The method has multiple neighborhood operators built in, each corresponding to a specific sequence adjustment logic. For example, there are three core operators: Lean Insert, which only changes the position of the urgent order in the sequence, used for fine adjustment; Ripple Swap, which exchanges the positions of any two regular orders after the urgent order without changing its position, used to optimize downstream planning; and Block Move, which moves a continuous segment (a block) of orders in the sequence to a new position, a more exploratory operation with stronger perturbation. The selection of the operator uses an adaptive strategy based on the search stage, that is, the selection probability of each operator is dynamically adjusted according to the current temperature. In the early stage of simulated annealing, the current temperature is high, and the algorithm needs stronger exploration ability, so the selection probability of high-perturbation operators such as Block Move is increased. In the later stage, the temperature decreases, and the algorithm tends to be stable, so the probability of fine adjustment operators such as Lean Insert is increased to perform local optimization. For example, the selection probability formula of the Lean Insert operator can be set as: P_LI(T)=P_{LI_min}+(P_{LI_max}-P_{LI_min})*(1-T / T_{initial}). Where P_{LI_min} and P_{LI_max} are the minimum and maximum probabilities, for example 0.2 and 0.7, T is the current temperature, and T_{initial} is the initial temperature. In this iteration, T and T_{initial} are both 100, so P_LI(100)=0.2. The remaining 0.8 probability is distributed to the other two operators according to the current high-temperature state, for example, set the Ripple Swap probability to 0.3 and the Block Move probability to 0.5. A random number between 0 and 1 is generated, and the operator used this time is determined according to the probability interval in which the random number falls. If the generated random number is 0.72, it falls within the probability interval of Block Move, so the neighborhood operator selected in this iteration is Block Move. After determining the operator, the random parameters required for its execution need to be generated. For the Block Move operator, three parameters need to be determined: block size k, block start index idx_start, and block new insertion position idx_insert. For the current sequence containing 4 orders, set the block size k to 2.A starting index is randomly selected, ensuring that the block does not go out of bounds, for example, idx_start is randomly selected as 1. At this point, the selected order block is the 2nd and 3rd order in the sequence, i.e., "Emergency Order JJ-20250926-001" and "Regular Order B". Next, a new insertion position is randomly selected, which cannot be inside the original block, for example, idx_insert is randomly selected as 4, representing moving the block to the end of the entire sequence. At this point, the dynamic neighborhood operator selection phase is complete, and its output is the selected neighborhood operator Block Move and its execution parameters {k = 2, idx_start = 1, idx_insert = 4}. Next, the neighborhood solution sequence construction phase is entered. First, a copy of the current solution sequence is created to avoid modifying the original state. Then, the copy is operated according to the operator and parameters. In this example, the order block with a starting index of 1 and a length of 2, i.e., "Emergency Order JJ-20250926-001" and "Regular Order B", is removed as a whole from the copy sequence. After removal, the copy sequence becomes: "Regular Order A remaining 70%" -> "Regular Order C". Subsequently, the order block just removed is inserted as a whole at index 4 of the adjusted sequence, i.e., the end. Finally, the constructed new sequence is: "Regular Order A remaining 70%" -> "Regular Order C" -> "Emergency Order JJ-20250926-001" -> "Regular Order B". This newly generated sequence is the neighborhood solution. In particular, these parameters, including the preset minimum / maximum probability, the specific probability of RippleSwap and Block Move, and the size of block movement, are all hyperparameters determined based on expert experience in solving cable scheduling problems and through repeated experiments and optimization on a large number of benchmark problems.

[0055] Next, the process in step 422 is entered to evaluate the new neighborhood solution. The impact cost function defined in step 33 is called to recalculate the total impact cost of the new neighborhood solution sequence as input. This calculation process is exactly the same as before, and the total changeover cost, weighted delay penalty, and plan disturbance penalty under this sequence are all taken into account. After calculation, it is found that the total impact cost of this new sequence is 45.0, which is recorded as the impact cost of the neighborhood solution.

[0056] After that, step 423 is entered to calculate the cost difference. The impact cost of the neighborhood solution 45.0 is compared with the cost 37.5 in the current solution state, and the cost difference between the two is calculated. The cost difference AC = 45.0 - 37.5 = 7.5. This positive cost difference indicates that the newly generated neighborhood solution is actually a worse solution than the current solution.

[0057] Then, the decision-making step 424 is entered, and the Metropolis acceptance criterion is applied. First, check whether the cost difference AC is less than 0. In this example, AC is 7.5, which is not less than 0, so this worse solution cannot be accepted directly. At this time, the logic of probabilistic acceptance is entered. In a feasible solution, step 424, if the cost difference is greater than or equal to zero, the probability of accepting the neighborhood solution is calculated and it is determined whether to update the current solution state to the combination of the new neighborhood solution and the impact cost of the neighborhood solution, including: the probability of the neighborhood solution is calculated as follows:

[0058]

[0059] wherein, AC is the cost difference, T is the current temperature, P is the probability of the neighborhood solution. This formula is the essence of the simulated annealing algorithm. The size of the probability P is inversely proportional to the cost difference, that is, the more the difference, the lower the acceptance probability, and is proportional to the current temperature, that is, the higher the temperature, the more inclined to accept the difference solution to explore a wider space. Substitute the current numerical value into the formula: P = e (-7.5 / 100) ≈ 0.927. This means that even if the cost of this new solution is 7.5 higher, there is still a probability of up to 92.7% to accept it in the current high-temperature state. Subsequently, a random floating point number between 0 and 1 is generated internally, for example, 0.65 is generated. Because 0.65 is less than the calculated acceptance probability 0.927, the result of this judgment is “accept”. Based on this result, the current solution state is updated, and the sequence therein is updated to the new neighborhood solution sequence “regular order A remaining 70%” -> “regular order C” -> “urgent order JJ-20250926-001” -> “regular order B”, and its cost is updated to 45.0. This updated current solution state will be used as the starting point for the next iteration.

[0060] Finally, independent of the above steps, step 425 is performed to update the optimal solution state, ensuring that the solution with the lowest cost in the entire search history is never lost. The impact cost 45.0 of the neighborhood solution is compared with the cost 37.5 in the current recorded optimal solution state. Since 45.0 is not less than 37.5, this neighborhood solution is not the best solution found so far. Therefore, the optimal solution state is not updated, and it still remains as {sequence: “regular order A...” -> “urgent order...”, cost: 37.5}.

[0061] After this series of operations, the single-iteration search and state update ends. The output is the updated current solution state {sequence: "Regular Order A remaining 70%" -> "Regular Order C" -> "Emergency Order JJ-20250926-001" -> "Regular Order B", cost: 45.0} and the unchanged optimal solution state {sequence: "Regular Order A remaining 70%" -> "Emergency Order JJ-20250926-001" -> "Regular Order B" -> "Regular Order C", cost: 37.5}.

[0062] Finally, step 43 is entered. The current temperature 100 from the previous iteration and the preset simulated annealing parameters {initial temperature: 100, termination temperature: 0.1, cooling rate: 0.95} are obtained. First, temperature update, i.e., annealing operation, is performed. This is the core link in the simulated annealing algorithm that simulates the physical cooling process. The current temperature will be reduced according to the preset cooling rate, and the calculation formula is: new temperature = current temperature * cooling rate. Substituting the specific numerical value, new temperature = 100 * 0.95 = 95. This new temperature value will be used for the next round of iterative search, and a lower temperature means that in the next round, the probability of accepting a poor solution will be reduced, making the search behavior of the algorithm more cautious, gradually shifting from extensive exploration to local optimization. Next, the termination condition is judged. The newly calculated temperature 95 is compared with the termination temperature 0.1 set in the simulated annealing parameters. The termination temperature is a threshold very close to zero. When the temperature is reduced to this level, it is considered that the algorithm has been fully searched and stabilized, and the benefit of continuing to search is very low. In this example, 95 is much greater than 0.1, so the termination condition is not met. Finally, according to the judgment result, the loop or termination control is performed. Since the termination condition is not met, the search process needs to continue. The updated current solution state and the updated optimal solution state output by the last iteration, together with the newly calculated temperature 95, are taken as a new set of input, and step 42, i.e., single-iteration search and state update, is called and executed again. This process from step 42 to step 43 and back to step 42 forms a complete iteration loop. This loop will continue. In the next iteration, the algorithm will start from the solution with a cost of 45.0, generate new neighborhood solutions under the condition of a temperature of 95, and make judgments. This cycle continues, and the temperature continues to decrease (95, 90.25, 85.74,...), while the optimal solution state always records the solution with the lowest cost encountered during the search process. After hundreds or even thousands of iterations, when the temperature is reduced to, for example, 0.09, the termination condition is met because 0.09 is less than or equal to 0.1 when the termination condition is judged next time. At this time, the loop will no longer continue, and the search process is officially over. The sequence of the solution with the lowest cost saved in the final updated optimal solution state is extracted as the final product of the entire simulated annealing search process, i.e., the optimal solution.

[0063] In particular, for the standard simulated annealing algorithm, its core lies in the fixed cooling strategy, that is, no matter whether the current search result is a large number of cost similar solutions or the cost changes sharply, the current temperature decays at a fixed cooling rate. This static strategy cannot perceive and respond to the local characteristics of the search space, resulting in low efficiency. Specifically, the current temperature is the only indicator to control the exploration ability of the algorithm, a higher temperature means that the algorithm has a high probability of accepting a higher cost difference solution, making it in a strong exploration mode, which is beneficial to jump out of the local optimal solution trap; while a lower temperature makes the algorithm almost only accept a lower cost good solution, making it in a strong utilization mode, which is beneficial to fine search around the current optimal solution. The ideal search process should be able to improve the exploration ability when encountering difficult areas (for example, falling into local optimum, the cost difference of neighborhood solutions is generally larger, resulting in a low acceptance rate); and reduce the exploration ability to speed up the convergence when in a flat area or find a promising descent path. Therefore, the optimization of the temperature update operation is introduced, which is to establish an adaptive cooling mechanism based on the acceptance rate, to dynamically adjust the current temperature by monitoring the actual acceptance rate of the difference solution in real time, so that the exploration and utilization rhythm of the algorithm can intelligently match the actual difficulty of the current search area. Through this feedback mechanism, the change of the current temperature is no longer a preset fixed path, but a dynamic adjustment according to the actual performance of the algorithm, realizing the self-adaptation to the characteristics of the search space.

[0064] In a feasible preferred solution, the optimization of the temperature update operation in the above step 43 includes: within the iteration window length, calculating the actual difference solution acceptance rate based on the reset difference solution acceptance count and the difference solution attempt count; obtaining a target difference solution acceptance rate, and comparing the actual difference solution acceptance rate with the target difference solution acceptance rate, in response to the actual difference solution acceptance rate being lower than the target difference solution acceptance rate, dividing the current temperature by a temperature adjustment factor to obtain an updated temperature; in response to the actual difference solution acceptance rate being higher than the target difference solution acceptance rate, multiplying the current temperature by the temperature adjustment factor to obtain the updated temperature.

[0065] Specifically, the implementation process of the optimization is to add an outer loop larger than the original simulated annealing iteration loop, and the outer loop has a preset iteration window length as a period, and the current temperature is adaptively adjusted at the end of each period.

[0066] Firstly, three key parameters need to be set: iteration window length, target difference solution acceptance rate and temperature adjustment factor. The iteration window length L is an integer, which defines the number of iterations for calculating the actual difference solution acceptance rate, such as setting it to about 100 times. The target difference solution acceptance rate target is a floating point number between 0 and 1, representing the expected exploration strength of the algorithm during the search process, and is set between 0.3 and 0.5. The temperature adjustment factor is a float number between 0 and 1, which is used to control the step size of temperature adjustment, and is set between 0.85 and 0.99. The values of these three key parameters are based on the empirical values in the related algorithm field, and are finally determined through experimental debugging for specific scheduling problems to ensure that the algorithm achieves the best balance between exploration efficiency and convergence speed.

[0067] First step: Calculate the actual worse solution acceptance rate based on the worse solution acceptance count and the worse solution trial count within the iteration window length. First, a quantitative indicator is needed to objectively evaluate the actual exploration behavior of the algorithm in the recent period of time, obtaining a feedback signal that accurately reflects the exploration intensity of the algorithm under the current temperature, i.e., the actual worse solution acceptance rate. In implementation, at the beginning of each iteration window period L, first initialize and reset two counters: the worse solution acceptance counter N_accepted_worse = 0 and the worse solution trial counter N_tried_worse = 0. Then, start the window loop, which is executed for the set number of times within the iteration window length. In each iteration, when a new neighborhood solution is generated and its cost is calculated, if it is found to be a worse solution, i.e., ), the worse solution trial counter is incremented by one. If this worse solution is further accepted probabilistically according to the Metropolis acceptance criterion, the worse solution acceptance counter is also incremented by one. After completing the iteration of the entire window, the actual worse solution acceptance rate within the window is calculated according to the following formula: actual = N_accepted_worse / N_tried_worse, where actual is the actual worse solution acceptance rate, N_accepted_worse is the worse solution acceptance counter, and N_tried_worse is the worse solution trial counter. That is, in the recent iteration window, the proportion of the final accepted worse solutions among all the tried worse solutions. This calculation result will be used as the basis for decision-making for subsequent temperature adjustment.

[0068] Second and third steps: Obtain the target worse solution acceptance rate, compare and update the temperature. Then, a decision-making mechanism is needed to determine whether the current temperature is too high or too low based on the feedback signal obtained in the first step, and to adjust it accordingly to achieve dynamic adaptive adjustment of the temperature, so that the algorithm can intelligently escape from local optimum and accelerate convergence. In implementation, compare the actual worse solution acceptance rate calculated in the previous step with the preset target worse solution acceptance rate. If the actual worse solution acceptance rate actual is lower than the target worse solution acceptance rate target, it indicates that the number of worse solutions accepted by the algorithm under the current temperature is too small, the exploration ability is insufficient, the algorithm behavior is too conservative, and there is a risk of falling into local optimum. To cope with this situation, the exploration ability needs to be improved, i.e., the temperature needs to be raised. At this time, the updated temperature is calculated according to the following formula: wherein is the updated temperature, is the current temperature, is the temperature adjustment factor. Since the temperature adjustment factor is a number less than 1, dividing the current temperature by it results in a larger value, thus achieving the effect of raising the temperature, enhancing the ability of the algorithm to accept worse solutions in the next iteration window. Conversely, if the actual worse solution acceptance rate actual is higher than the target worse solution acceptance rate target, it indicates that the algorithm is accepting too many worse solutions at the current temperature, and the exploration behavior is too random, which is not conducive to stable convergence to the optimal solution. To cope with this situation, the exploration ability needs to be reduced to speed up the convergence process, that is, the cooling operation is performed. At this time, the method calculates the updated temperature according to the following formula: . Multiplying the current temperature by a factor less than 1 results in a smaller value, thus achieving the effect of lowering the temperature, making the algorithm's behavior in the next iteration window more inclined to exploitation and convergence.

[0069] In this way, the acceptance rate-based adaptive cooling mechanism can significantly reduce the sensitivity of the initial temperature and cooling parameters, and even imperfect initial temperatures can be quickly adjusted to the appropriate range within the first few windows, and users no longer need to struggle to debug fixed cooling rates. Moreover, when the algorithm falls into a deep local optimum, almost all neighborhood solutions are worse solutions with larger cost differences, which will cause the actual acceptance rate actual to drop sharply, and the adaptive mechanism will react by dividing (warming up) to significantly improve the algorithm's ability to escape from this trap, improving search efficiency and solution quality. At the same time, when the algorithm finds a path to the global optimum, most neighborhood solutions are good solutions, and the acceptance rate of worse solutions will naturally remain at a high level. At this time, the adaptive mechanism will also speed up the convergence by multiplying (cooling down) to avoid unnecessary waste of time on excessive random exploration.

[0070] In step 5, the optimal solution is converted into a new set of scheduling instructions that the MES system can execute. Accordingly, after a complex metaheuristic iterative search, a theoretically optimal scheduling scheme has been obtained. However, this scheme itself is only a set of data structures within the algorithm, which describes an ideal order sequence and its cost, but cannot directly drive the physical equipment of the factory for production. In modern cable manufacturing workshops, the Manufacturing Execution Module (MES) is the core hub connecting the planning layer and the execution layer, which is responsible for issuing production instructions to specific production lines and workstations. In order to make the optimization results of the algorithm truly land, it is necessary to convert this abstract optimal solution into standardized and formatted instructions that the MES can understand and execute. Therefore, the present application finally converts the optimal solution into a new set of scheduling instructions that the MES system can execute, to seamlessly integrate the optimization results into the actual production process, thereby ensuring that the dynamic embedding of urgent orders is accurately and efficiently completed.

[0071] In one implementation, step 5 operates as follows: After thousands of iterations, the algorithm may have found a solution with a lower cost than the initial solution. For example, the final optimal solution is: a production sequence of "regular order A remaining 70%" -> "emergency order JJ-20250926-001" -> "regular order C" -> "regular order B" with an associated total impact cost of 35.2. This optimal solution not only contains the order sequence, but also includes the exact planned start and finish times for each order calculated based on the sequence.

[0072] The first step is to present this optimal solution data to the production scheduler for final confirmation. The optimal solution sequence and its detailed timing are rendered into an intuitive and understandable user interface view, such as a Gantt chart or a detailed production schedule table. This table clearly lists the details of each production task, including the process sequence, order number, product specification, planned start time, planned finish time, and estimated changeover time. For example, the table seen by the scheduler may show: the remaining part of regular order A starts at 10:00 and finishes at 19:30; followed by a 1.5-hour changeover; then the emergency order starts production at 21:00 and finishes at 05:00 the next day; after a 0.6-hour changeover, regular order C starts production at 05:36, and so on. This manual review step is an important guarantee to ensure the feasibility of the solution in actual operation, allowing the scheduler to make final decisions based on some implicit knowledge not covered by the algorithm.

[0073] After the scheduler confirms the solution, the second step is to convert this confirmed optimal solution into a formatted instruction set that the MES can directly parse and execute, i.e., a new scheduling instruction set. This instruction set is a standardized data format, such as JSON or XML. According to the pre-defined data structure template, instructions are generated one by one. For example, a JSON-formatted instruction set may be an array containing multiple task objects. For each order in the optimal solution, a corresponding task object is generated, detailing its execution parameters. For example, the task object generated for the emergency order may be as follows: { "task ID": "TSK-20250926-102", "order number": "JJ-20250926-001", "product code": "YJV-1kV-4x185", "production quantity": 500, "unit": "meters", "equipment number": "L1", "planned start timestamp": "2025-09-26T21:00:00Z", "planned end timestamp": "2025-09-27T05:00:00Z", "predecessor task changeover time": 5400 seconds}. Similar structured instructions are generated for all four orders in the sequence, collectively forming the complete new scheduling instruction set.

[0074] The last step is to issue the generated new scheduling instruction set to the MES through a preset interface. This process is completed through a secure network request. An interface function is called to send the request containing the above-mentioned JSON instruction set to the receiving address specified by the MES. After the MES receives the instruction set, it will first parse and verify it to confirm the compliance and integrity of the instructions. After verification, the MES will overwrite or update the original task queue of production line L1 in its database with the new task sequence. At this point, the production plan of the workshop is officially and dynamically updated. The MES will issue production instructions to the control unit and operation terminal of production line L1 in sequence and according to the time arrangement in the new instruction set, so as to accurately execute the new production scheduling scheme optimized globally and complete the embedding of the urgent insertion order.

[0075] In summary, the intelligent production scheduling method for cable flexible manufacturing based on the embodiments of the present application is illustrated, which solves the problem of dynamic embedding and minimum influence of urgent insertion orders in cable flexible manufacturing, and adopts a two-stage strategy of fast initialization + global optimization. First, through data solidification, dynamic information such as urgent orders, current plans, factory status and changeover time is comprehensively captured to form the context basis for decision-making. Then, a fast heuristic algorithm is used to quantitatively evaluate the cost of all possible insertion schemes of the urgent order, and a high-quality initial scheduling solution is quickly generated to avoid blind search. Finally, a simulated annealing meta-heuristic algorithm is introduced to perform global iterative search starting from the initial solution, and the probabilistic escape mechanism can effectively avoid the local optimal trap that the traditional method is prone to, so as to find the optimal solution with the minimum disturbance to the overall production plan. This method converts complex experience-based decision-making into a data-driven automated optimization process, realizes rapid and accurate adjustment of the production plan, and improves the overall stability and efficiency of the production line.

[0076] Figure 5 The block diagram of the intelligent production scheduling system for cable flexible manufacturing according to the embodiments of the present application is shown in FIG. 1. As shown in FIG. 1, the system comprises a data acquisition module 101, a data preprocessing module 102, a fast heuristic algorithm module 103, a simulated annealing meta-heuristic algorithm module 104, a data storage module 105, a data analysis module 106, and a data output module 107. Figure 5As shown, the intelligent production scheduling system 100 for cable flexible manufacturing according to the embodiment of the present application comprises: a production data acquisition module 110, configured to acquire emergency order data, current production plan, real-time factory state and changeover time matrix; a data solidification module 120, configured to perform data solidification on the emergency order data, the current production plan, the real-time factory state and the changeover time matrix to obtain problem context; an initial solution generation module 130, configured to perform fast heuristic initialization on the emergency order data and the current production plan based on the problem context to obtain an initial solution; an optimal solution generation module 140, configured to perform simulated annealing-based meta-heuristic iterative search on the initial solution based on simulated annealing parameters to obtain an optimal solution, the simulated annealing parameters comprising initial temperature, termination temperature and cooling rate; and a scheduling module 150, configured to convert the optimal solution into a new scheduling instruction set executable by the MES system.

[0077] Here, those skilled in the art can understand that the specific operations of each step in the above intelligent production scheduling system for cable flexible manufacturing have been described in detail above with reference to the intelligent production scheduling method for cable flexible manufacturing, and therefore repeated descriptions thereof will be omitted. Figures 1 to 4 Here, those skilled in the art can understand that the specific operations of each step in the above intelligent production scheduling system for cable flexible manufacturing have been described in detail above with reference to the intelligent production scheduling method for cable flexible manufacturing, and therefore repeated descriptions thereof will be omitted.

Claims

1. A method for intelligent production scheduling for cable flexible manufacturing, characterized in that, The method comprises: acquiring emergency order data, current production plan, real-time factory status and changeover time matrix; data solidification is performed on the emergency order data, current production plan, real-time factory status and changeover time matrix to obtain problem context; based on the problem context, quick heuristic initialization is performed on the emergency order data and current production plan to obtain an initial solution; based on simulated annealing parameters, meta-heuristic iterative search is performed on the initial solution based on simulated annealing to obtain an optimal solution, the simulated annealing parameters including initial temperature, termination temperature and cooling rate; the optimal solution is converted into a new set of scheduling instructions executable by the MES system.

2. The intelligent production scheduling method for cable-oriented flexible manufacturing according to claim 1, characterized in that, The quick heuristic initialization performed on the emergency order data and current production plan based on the problem context to obtain an initial solution comprises: generating n+1 logical insertion points based on the length n of the current production plan; inserting the emergency order data into each of the logical insertion points to obtain a candidate plan sequence list; quantifying the impact cost of each candidate plan in the candidate plan sequence list based on the problem context to obtain a scored candidate plan list; performing minimum cost search on the scored candidate plan list to obtain an initial solution.

3. The intelligent production scheduling method for cable-oriented flexible manufacturing according to claim 2, characterized in that, The impact cost of each candidate plan includes weighted tardiness penalty, total changeover cost and plan disturbance penalty.

4. The intelligent production scheduling method for cable-oriented flexible manufacturing according to claim 3, characterized in that, Wherein, the weighted tardiness penalty is calculated by the following formula: wherein, is a candidate schedule, is a candidate schedule in a list of candidate schedules, is an order in the candidate schedule, is an order in the candidate schedule, is an order completion time for the order, is an order priority for the order, is an order due date for the order, is an order due date for the order, is an order due date for the order, is an order due date for the order, is an order due date for the order, is a max function, is a weighted tardiness penalty for the candidate schedule, is a weighted tardiness penalty for the candidate schedule. the total changeover cost is calculated by the following formula: wherein, is a function of the changeover time, is the current order for the candidate plan, is the preceding order in the candidate plan, is the preceding order in the candidate plan, is the total changeover cost for the candidate plan; the plan disturbance penalty is calculated by the following formula: wherein, is the number of affected orders, is the average delay time, is the candidate schedule a schedule perturbation penalty.

5. The cable-flexible manufacturing oriented smart production scheduling method of claim 1, wherein, The meta-heuristic iterative search performed on the initial solution based on simulated annealing to obtain an optimal solution comprises: search state initialization based on the simulated annealing parameters and the initial solution to obtain a current solution state, an optimal solution state and a current temperature; single-iteration search and state update of the current solution state, the optimal solution state and the current temperature based on a neighborhood generating function and an impact cost function to obtain an updated current solution state and an updated optimal solution state; temperature annealing and loop control of the updated current solution state and the updated optimal solution state to obtain an optimal solution.

6. The cable-flexible manufacturing oriented smart production scheduling method according to claim 5, wherein, The single-iteration search and state update of the current solution state, the optimal solution state and the current temperature based on a neighborhood generating function and an impact cost function to obtain an updated current solution state and an updated optimal solution state comprises: inputting the current solution state into the neighborhood generating function to obtain a new neighborhood solution; inputting the new neighborhood solution into the impact cost function to obtain the impact cost of the neighborhood solution; calculating the cost difference between the impact cost of the neighborhood solution and the impact cost of the current solution; if the cost difference is less than 0, directly updating the current solution state to the combination of the new neighborhood solution and the impact cost of the neighborhood solution; if the cost difference is greater than or equal to zero, calculating the probability of accepting the neighborhood solution and determining whether to update the current solution state to the combination of the new neighborhood solution and the impact cost of the neighborhood solution; comparing the impact cost of the neighborhood solution with the impact cost of the optimal solution, and if the impact cost of the neighborhood solution is less than the impact cost of the optimal solution, updating the optimal solution state to the combination of the new neighborhood solution and the impact cost of the neighborhood solution.

7. The cable-flexible manufacturing oriented smart production scheduling method according to claim 6, characterized in that, If the cost difference is greater than or equal to zero, a probability of accepting the neighbor solution is calculated and it is determined whether to update the current solution state to a combination of the new neighbor solution and the influence cost of the neighbor solution, comprising: calculating the probability of the neighbor solution by the following formula: wherein, is the cost difference, is the current temperature, is the probability of the neighborhood solution.

8. An intelligent production scheduling system for cable-flexible manufacturing, characterized in that, Comprising: A production data acquisition module configured to acquire emergency order data, a current production plan, real-time factory status, and a model changeover time matrix; A data solidification module configured to perform data solidification on the emergency order data, the current production plan, the real-time factory status, and the model changeover time matrix to obtain a problem context; An initial solution generation module configured to perform a quick heuristic initialization on the emergency order data and the current production plan based on the problem context to obtain an initial solution; An optimal solution generation module configured to perform a simulated annealing-based meta-heuristic iterative search on the initial solution based on simulated annealing parameters to obtain an optimal solution, the simulated annealing parameters including an initial temperature, a termination temperature, and a cooling rate; A scheduling module configured to convert the optimal solution into a new set of scheduling instructions executable by an MES system.

Images