Factory allocation method for distributed job shop scheduling
By introducing the partition number theory and cosine similarity calculation method, the factory allocation of distributed job shop scheduling is optimized, the problem of increased maximum completion time caused by job allocation is solved, and load balancing between factories and improved scheduling efficiency are achieved.
Patent Information
- Application Number
- CN202510955384.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-11
- Publication Date
- 2025-09-26
AI Technical Summary
Existing greedy heuristic algorithms cannot effectively quantify the similarity of process routes between workpieces in distributed job shop scheduling problems, resulting in concentrated allocation of workpieces to the same factory, increasing the maximum completion time and reducing scheduling efficiency.
The partition number theory and cosine similarity calculation method are introduced to optimize the factory allocation strategy by quantifying the operation priority relationship and process route similarity between workpieces, avoiding the allocation of highly similar workpieces to the same factory, and reducing the solution space complexity and resource competition.
Effectively reduce the scale of the solution space, optimize the load balance between factories, improve scheduling efficiency and solution efficiency, reduce competition for machine resources, and improve the balance of maximum completion time.
Smart Images
Figure CN120706831A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of production scheduling, and in particular to a factory allocation method for distributed job shop scheduling. Background Art
[0002] The distributed job shop scheduling problem is an extension of the traditional job shop scheduling problem, extending the problem from a single factory to multiple factories with identical configurations. Workpieces must be assigned to a specific factory for processing, and once assigned, all processing steps must be completed within that factory. The goal of the distributed job shop scheduling problem is to optimize overall production efficiency and resource utilization by coordinating production activities across multiple factories.
[0003] Greedy heuristic algorithms are widely used for solving distributed job shop scheduling problems due to their high efficiency. Among them, allocation rules based on workload and maximum makespan are a common strategy. Specifically, workload-based rules assign workpieces to factories sequentially, prioritizing factories with the lowest machine loads after assignment. This balances workloads across factories and prevents some factories from becoming production bottlenecks due to the concentrated allocation of highly loaded workpieces. In contrast, maximum makespan-based rules assign workpieces to factories one by one in descending order of their total processing time, while simultaneously sorting the processes, fully accounting for the interactive effects of allocation and sorting. However, these rules have limitations: the core of the greedy heuristic algorithm prioritizes local optimal decisions, selecting from the optimal allocation or sorting scheme for each job at each step. However, this local greedy strategy often fails to guarantee a global optimal solution. It lacks a systematic analysis of the global solution space and does not directly quantify the similarity of process routes between workpieces. This can lead to workpieces with highly similar process routes being concentrated in the same factory, exacerbating competition for machine resources, extending maximum makespan, and reducing scheduling efficiency.
[0004] To address the issue of increased maximum completion time due to workpiece allocation, the insertion-based greedy heuristic algorithm considers the interaction between job-factory allocation and operation sequencing, taking into account the immediate effects of sequencing during allocation, thereby optimizing the overall scheduling results. Based on this, the existing technology proposes the GH3 allocation strategy, which uses an insertion-based search strategy. Each time a workpiece is inserted into an operation, all possible insertion locations are systematically evaluated and the location with the lowest maximum completion time is selected. This greedy strategy ensures that each workpiece allocation is locally optimal.
[0005] In the distributed job shop scheduling problem, the greedy heuristic algorithm GH3 achieves load balancing across factories by minimizing the maximum completion time of each factory. Specifically, the algorithm generates a clear allocation plan for workpieces and their process sequence within each factory through dynamic job-to-factory allocation and operation sequencing, thereby forming a deterministic feasible solution. However, this solution generation process does not directly quantify the feasible solution space and process route similarity, resulting in an inability to accurately assess the complexity of the search space. Workpieces with similar process routes are assigned to the same factory, resulting in increased competition for machine resources, increased maximum completion time, and reduced scheduling efficiency. Summary of the Invention
[0006] In order to solve the technical defects of distributed job shop scheduling problems, factory allocation search complexity and solution efficiency, the present invention introduces the precise measurement of the solution space by the partition number theory and the mathematical analysis of the process route by cosine similarity, and combines the professional knowledge in the field of distributed job shops to propose a novel factory allocation method. While achieving factory workload balancing, this method effectively avoids the problem of solution space limitation caused by premature detailed scheduling, reduces search complexity and improves solution efficiency.
[0007] In one aspect, to achieve the above-mentioned objectives, the present invention provides a factory allocation method for distributed job shop scheduling, comprising:
[0008] S1. Obtain a distributed job shop scheduling problem instance and determine the workpiece number, process route, and processing time;
[0009] S2. generating a random workpiece sequence based on the workpiece number;
[0010] S3. Based on the process route, calculating the number of divisions between the workpieces and the similarity of the process route;
[0011] S4, allocating the first N workpieces of the workpiece sequence to N factories for initial allocation;
[0012] S5, iteratively select the factory with the least number of assigned workpieces as the candidate factory;
[0013] S6. Calculate the number of splits and process route similarity of the workpiece to be assigned after it is added to each candidate factory;
[0014] S7. Select the candidate factory with the largest number of splits to allocate the workpiece. If the number of splits is the same, select the factory with the lowest process route similarity to allocate the workpiece.
[0015] S8. Repeat S5-S7 until all workpieces are allocated.
[0016] Preferably, generating a random artifact sequence comprises:
[0017] The workpiece numbers are arranged by a random algorithm to generate the random workpiece sequence, thereby avoiding specific rule sorting and enhancing the diversity of solutions.
[0018] Preferably, the calculation method of the split number is:
[0019]
[0020] Where, PREC j (p,q) represents the order in which job j occupies machine p and machine q in the process constraint; PREC k (p,q)) represents the order in which workpiece k occupies machine p and machine q in the process constraint; XOR is the logical exclusive OR operator. If XOR(PREC j (p,q),PREC k (p,q))=1, then there is a partition number between workpiece j and workpiece k between machine p and machine q; j and k are the index numbers of the workpieces, n and m are the total number of workpieces and the total number of machines respectively, and T is the total number of partitions of each workpiece.
[0021] Preferably, the process route similarity is measured using cosine similarity, specifically:
[0022]
[0023] Where, S(R i ,R j ) is the process route similarity; R i ,R j are the process routes of workpiece i and workpiece j respectively; p i,k' represents the processing time of workpiece i in the k'th process in the process route; δ(M i,k' ,M j,k' ) is the machine matching function; p j,k' represents the processing time of workpiece j in the k'th process in the process route; i and j are the index numbers of the workpiece; k' is the index number of the workpiece process; m is the total number of machines; S is the process route similarity between the two workpieces.
[0024] Preferably, determining the candidate factory includes:
[0025] If the number of workpieces allocated to multiple factories is the same minimum value, then all factories are candidate factories;
[0026] If there is only one factory with the least number of workpieces, the factory with the least number of workpieces is the candidate factory.
[0027] Preferably, the priorities for performing the iterative selection include:
[0028] First-level indicator: maximize the number of new splits of candidate factories;
[0029] Second-level indicator: Minimize the similarity of new process routes of candidate factories.
[0030] On the other hand, to achieve the above-mentioned object, the present invention further provides a factory allocation system for distributed job shop scheduling, comprising:
[0031] Data acquisition module: used to obtain workpiece number, process route and processing time;
[0032] Random sequence generator: used for generating a random artifact sequence based on the artifact number;
[0033] A division number and similarity calculation unit: used for calculating the division number and process route similarity between workpieces based on the process route;
[0034] Intelligent allocation module: used to make allocation decisions based on the number of splits and the similarity of the process routes;
[0035] Load monitoring module: used to track machine utilization and workpiece backlog in each factory in real time.
[0036] Compared with the prior art, the present invention has the following advantages and technical effects:
[0037] (1) This paper introduces the theory of partitioning numbers for the first time, which is used to quantify the diversity of operational priority relationships between workpieces. The partitioning number reflects the number of potential loops in the order in which workpieces are used in machines. A larger partitioning number indicates a higher probability of loops between workpieces, a smaller number of feasible solutions, and a smaller solution space. By prioritizing allocation solutions with large partitioning numbers, the solution space size of subproblems within the factory is effectively constrained at the initial allocation stage, reducing search complexity.
[0038] (2) This paper proposes a method for calculating process route similarity based on cosine similarity, which comprehensively considers the differences in machine usage sequence and processing time between workpieces and accurately quantifies the degree of process route overlap between workpieces. This method not only considers the overlap in machine usage but also incorporates differences in processing time, providing a more accurate similarity measure. Compared with traditional methods that do not directly quantify process route similarity, this method reduces machine resource competition and optimizes load balancing between factories by avoiding the allocation of highly similar workpieces to the same factory.
[0039] (3) The present invention combines the number of splits and the process route similarity to form a factory allocation framework. The number of splits is used as an indicator of solution space constraints, and the allocation scheme that can reduce the solution space is given priority; the process route similarity is used as an indicator of load balancing, which guides the allocation of workpieces to disperse resource competition. The two work together to not only reduce the scale of the solution space, but also ensure the balanced distribution of workload. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:
[0041] Figure 1 This is a flow chart of a factory allocation method for distributed job shop scheduling according to an embodiment of the present invention;
[0042] Figure 2 Detailed flowchart of an embodiment of the present invention;
[0043] Figure 3 A schematic diagram of a detailed workpiece allocation process according to an embodiment of the present invention;
[0044] Figure 4 This is an overall flow chart of the factory allocation method for distributed job shop scheduling according to an embodiment of the present invention. DETAILED DESCRIPTION
[0045] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0046] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.
[0047] like Figures 1-4 This embodiment proposes a factory allocation method for distributed job shop scheduling, including:
[0048] S1. Obtain a distributed job shop scheduling problem instance and determine the workpiece number, process route, and processing time;
[0049] S2, generating a random workpiece sequence based on the workpiece number,
[0050] S3. Based on the process route, calculating the number of divisions between the workpieces and the similarity of the process route;
[0051] S4, allocating the first N workpieces of the workpiece sequence to N factories for initial allocation;
[0052] S5, iteratively select the factory with the least number of assigned workpieces as the candidate factory;
[0053] S6. Calculate the number of splits and process route similarity of the workpiece to be assigned after it is added to each candidate factory;
[0054] S7. Select the candidate factory with the largest number of splits to allocate the workpiece. If the number of splits is the same, select the factory with the lowest process route similarity to allocate the workpiece.
[0055] S8. Repeat S5-S7 until all workpieces are allocated.
[0056] Furthermore, a random artifact sequence is generated, including:
[0057] The workpiece numbers are arranged by a random algorithm to generate the random workpiece sequence, thereby avoiding specific rule sorting and enhancing the diversity of solutions.
[0058] Furthermore, the calculation method of the number of divisions is:
[0059]
[0060] Where, PREC j (p,q) represents the order in which job j occupies machine p and machine q in the process constraint; PREC k (p,q)) represents the order in which workpiece k occupies machine p and machine q in the process constraint; XOR is the logical exclusive OR operator. If XOR(PREC j (p,q),PREC k (p,q))=1, then there is a partition number between workpiece j and workpiece k between machine p and machine q; j and k are the index numbers of the workpieces, n and m are the total number of workpieces and the total number of machines respectively, and T is the total number of partitions of each workpiece.
[0061] Specifically, the significance of the partition number T lies in quantifying the size of the solution space, especially when the scheduling solution is undetermined or uncertain. The partition number between workpieces is calculated to measure the number of potential loops between any two workpieces. The size of the partition number directly affects the complexity of the solution space. A larger partition number increases the number of potential priority constraints between workpieces, which reduces the number of feasible solutions and the size of the solution space, thereby reducing the difficulty of solving the scheduling problem. By prioritizing allocation solutions with a high partition number, the solution space size can be significantly compressed during the allocation process, reducing the computational burden of the subsequent task sorting stage.
[0062] Furthermore, the process route similarity is measured using cosine similarity. The significance of the partition number T is to quantify the scale of the solution space. Especially when the scheduling plan has not yet been determined or is in an uncertain state, the partition number reflects the diversity of the operation priority relationship under the process constraints between workpieces. It is specifically defined as follows:
[0063]
[0064] Where, S(R i ,R j ) is the process route similarity; Ri ,R j are the process routes of workpiece i and workpiece j respectively; p i,k' represents the processing time of workpiece i in the k'th process in the process route; δ(M i,k' ,M j,k' ) is the machine matching function; p j,k' represents the processing time of workpiece j in the k'th process in the process route; i and j are the index numbers of the workpiece; k' is the index number of the workpiece process; m is the total number of machines; S is the process route similarity between the two workpieces.
[0065] If workpiece i and workpiece j use the same machine in the k'th process, then δ(M i,k' ,M j,k' )==1; otherwise, δ(M i,k' ,M j,k' )=0. In this embodiment, S(R i ,R j ) is close to 1, indicating that the two workpieces are highly consistent in terms of machine usage sequence and processing time, and may compete for the same machine resources in scheduling. i ,R j ) is close to 0, indicating that the process routes of the two workpieces are significantly different and the resource competition is relatively small.
[0066] If workpieces with highly similar process routes are assigned to the same factory, it may lead to concentrated machine loads and increase the maximum completion time; while assigning workpieces with greatly different process routes to the same factory will help balance machine loads and optimize overall scheduling efficiency. i ,R j ) Process route similarity is the key basis for allocation decision-making. By distributing highly similar workpieces to different factories, it can effectively alleviate resource competition, reduce scheduling bottlenecks, and achieve balanced distribution of workloads among factories.
[0067] Taking full account of the constraints of machine usage sequence and processing time in the distributed job shop scheduling problem, the process route similarity S(R i ,R j ) quantifies the similarity between the process routes of workpieces i and j in terms of machine usage and processing time using the cosine similarity metric. When allocating jobs, avoid assigning jobs with high similarity to the same factory to reduce machine competition and balance the workload.
[0068] Furthermore, determining the candidate factories includes:
[0069] If the number of workpieces allocated to multiple factories is the same minimum value, then all factories are candidate factories;
[0070] If there is only one factory with the least number of workpieces, the factory with the least number of workpieces is the candidate factory.
[0071] Specifically, the partition number serves as a key quantitative indicator of solution space complexity. The larger its value, the greater the number of potential loops of operational priority relationships between workpieces. Specifically, when arbitrarily arranging the processing order of workpieces on each machine, a higher partition number tends to lead to loop formation, significantly reducing the number of feasible solutions, shrinking the size of the solution space, and lowering the complexity of solving the scheduling problem. While taking into account workload balancing between factories, the partition number is introduced as an optimization metric, aiming to reduce the scheduling complexity of each factory's subproblems through refined allocation strategies. By using the partition number as a guide for workpiece allocation, the size of the solution space of each factory can be effectively constrained, thereby significantly reducing the solution complexity, promoting efficient search within the local solution space, and thus improving the efficiency of finding the optimal solution. By optimizing the partition strategy, this method not only effectively controls the complexity of the subproblems but also provides an efficient solution space exploration framework for distributed job shop scheduling.
[0072] In the distributed job shop scheduling problem, a factory allocation framework is constructed by integrating the split number and process route similarity, aiming to synergistically improve the efficiency of solution space exploration and workload balance across factories. The split number accurately quantifies the diversity of operation priority relationships between workpieces and serves as the core indicator for constraining the scale of the solution space. Allocation schemes with higher split numbers are prioritized, effectively compressing the sorted solution space within each factory and accelerating the search efficiency in high-potential solution areas. Process route similarity is based on the cosine similarity metric of processing time and machine allocation. It accurately captures the degree of overlap between workpieces in machine usage sequence and processing time, covering the processing sequence and processing time of workpieces on the machine. It disperses highly similar workpieces to different factories, effectively alleviating machine resource competition and significantly reducing scheduling bottlenecks caused by resource concentration, thereby optimizing the balanced distribution of workload across factories.
[0073] This embodiment also provides a factory allocation system for distributed job shop scheduling, including:
[0074] Data acquisition module: used to obtain workpiece number, process route and processing time;
[0075] Random sequence generator: used for generating a random artifact sequence based on the artifact number;
[0076] A division number and similarity calculation unit: used for calculating the division number and process route similarity between workpieces based on the process route;
[0077] Intelligent allocation module: used to make allocation decisions based on the number of splits and the similarity of the process routes;
[0078] Load monitoring module: used to track machine utilization and workpiece backlog in each factory in real time.
[0079] In order to more clearly express the technical solution of the present invention, the following specific embodiments are provided to introduce the solution:
[0080] This example uses the classic FT10 example, which includes 10 workpieces and 10 machines, with a total of 100 processes. Table 1 lists the processing machines and processing time for each process of the workpiece. Table 2 lists the number of divisions between workpieces, and Table 3 lists the similarity of process routes between workpieces. The following are the specific steps for allocating workpieces to factories, as shown in the following example: Figure 4 :
[0081] Step 1: According to the FT10 example, randomly arrange the workpieces. The initial workpiece order is {4, 3, 9, 6, 8, 2, 10, 1, 7, 5}.
[0082] Step 2: Assign the first three workpieces 4, 3, and 9 in the workpiece sorting to factories 1, 2, and 3 respectively. At this point, one workpiece is assigned to each factory, completing the initial assignment. The candidate factories are now 1, 2, and 3.
[0083] Step 3: Based on the FT10 case information, the number of splits and process route similarity of each workpiece are calculated as shown in Tables 2 and 3.
[0084] Step 4: Assign the fourth workpiece (workpiece 6) to each candidate factory and calculate the number of splits for each candidate factory. The calculated splits for factories 1, 2, and 3 are 21, 13, and 15, respectively. Therefore, workpiece 6 is assigned to factory 1.
[0085] Step 5: Assign the fifth workpiece in the sorted order (workpiece 8). Workpiece 8 is assigned to the candidate factories (factories 2 and 3). The calculated partition counts for factories 2 and 3 are 18 and 16, respectively. Therefore, workpiece 8 is ultimately assigned to factory 2.
[0086] Step 6: Assign the sixth workpiece (workpiece 2) in the workpiece sorting. Since there are only three candidate factories at this point, only factory 3 has the fewest assigned workpieces. Therefore, workpiece 2 is assigned to factory 3.
[0087] Step 7: After assigning the first six workpieces in the workpiece sorting, the candidate factories are now Factory 1, Factory 2, and Factory 3. Next, assign the seventh workpiece in the workpiece sorting (workpiece 10). Try to assign workpiece 7 to each candidate factory and calculate the number of splits for each candidate factory. The calculated number of splits for Factories 1, 2, and 3 is 46, 40, and 46, respectively. The factories with the largest number of splits are Factories 1 and 3. At this time, the process route similarities for Factories 1 and 3 should be calculated to be 0.57 and 0.283, respectively. Therefore, assign workpiece 10 to Factory 3.
[0088] Step 8: Repeat the above steps until all workpieces are assigned to factories. The final workpiece assignment result is: workpieces 4, 6, and 1 are assigned to factory 1, workpieces 3, 8, and 7 are assigned to factory 2, and workpieces 9, 2, 10, and 5 are assigned to factory 3.
[0089] Table 1
[0090]
[0091] Table 2
[0092] Workpiece number J1 J2 J3 J4 J5 J6 J7 J8 J9 J10 J1 0 11 11 11 8 16 11 11 9 14 J2 11 0 22 16 15 21 16 14 14 17 J3 11 22 0 16 11 13 6 18 10 11 J4 11 16 16 0 15 21 16 10 20 11 J5 8 15 11 15 0 10 15 9 13 16 J6 16 21 13 21 10 0 13 15 15 14 J7 11 16 6 16 15 13 0 16 6 9 J8 11 14 18 10 9 15 16 0 16 11 J9 9 14 10 20 13 15 6 16 0 15 J10 14 17 11 11 16 14 9 11 15 0
[0093] Table 3
[0094] Workpiece number J1 J2 J3 J4 J5 J6 J7 J8 J9 J10 J1 1.000 0.162 0.004 0.000 0.134 0.137 0.011 0.059 0.517 0.020 J2 0.162 1.000 0.000 0.210 0.070 0.000 0.083 0.078 0.201 0.000 J3 0.004 0.000 1.000 0.200 0.127 0.270 0.421 0.191 0.071 0.376 J4 0.000 0.210 0.200 1.000 0.048 0.000 0.119 0.000 0.000 0.276 J5 0.134 0.070 0.127 0.048 1.000 0.045 0.010 0.256 0.254 0.003 J6 0.138 0.000 0.270 0.000 0.045 1.000 0.000 0.076 0.029 0.294 J7 0.011 0.083 0.421 0.119 0.010 0.000 1.000 0.155 0.154 0.157 J8 0.059 0.078 0.191 0.000 0.256 0.076 0.155 1.000 0.099 0.032 J9 0.517 0.201 0.071 0.000 0.254 0.029 0.154 0.099 1.000 0.082 J10 0.020 0.000 0.376 0.276 0.003 0.294 0.157 0.032 0.082 1.000
[0095] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.
Claims
1. A factory allocation method for distributed job shop scheduling, characterized in that: include: S1. Obtain a distributed job shop scheduling problem instance and determine the workpiece number, process route, and processing time; S2. generating a random workpiece sequence based on the workpiece number; S3. Based on the process route, calculating the number of divisions between the workpieces and the similarity of the process route; S4, allocating the first N workpieces of the workpiece sequence to N factories for initial allocation; S5, iteratively select the factory with the least number of assigned workpieces as the candidate factory; S6. Calculate the number of splits and process route similarity of the workpiece to be assigned after it is added to each candidate factory; S7. Select the candidate factory with the largest number of splits to allocate the workpiece. If the number of splits is the same, select the factory with the lowest process route similarity to allocate the workpiece. S8. Repeat S5-S7 until all workpieces are allocated.
2. The factory allocation method for distributed job shop scheduling according to claim 1, characterized in that: Generates a random sequence of artifacts consisting of: The workpiece numbers are arranged by a random algorithm to generate the random workpiece sequence, thereby avoiding specific rule sorting and enhancing the diversity of solutions.
3. The factory allocation method for distributed job shop scheduling according to claim 1, characterized in that: The calculation method of the split number is: Where, PREC j (p,q) represents the order in which job j occupies machine p and machine q in the process constraint; PREC k (p,q)) represents the order in which workpiece k occupies machine p and machine q in the process constraint; XOR is the logical exclusive OR operator. If XOR(PREC j (p,q),PREC k (p,q))=1, then there is a partition number between workpiece j and workpiece k between machine p and machine q; j and k are the index numbers of the workpieces, n and m are the total number of workpieces and the total number of machines respectively, and T is the total number of partitions of each workpiece.
4. The factory allocation method for distributed job shop scheduling according to claim 1, characterized in that: The process route similarity is measured using cosine similarity, specifically: Where, S(R i ,R j ) is the process route similarity; R i ,R j are the process routes of workpiece i and workpiece j respectively; p i,k' represents the processing time of workpiece i in the k'th process in the process route; δ(M i,k' ,M j,k' ) is the machine matching function; p j,k' represents the processing time of workpiece j in the k'th process in the process route; i and j are the index numbers of the workpiece; k' is the index number of the workpiece process; m is the total number of machines; S is the process route similarity between the two workpieces.
5. The factory allocation method for distributed job shop scheduling according to claim 1, characterized in that: The candidate factories identified include: If the number of workpieces allocated to multiple factories is the same minimum value, then all factories are candidate factories; If there is only one factory with the least number of workpieces, the factory with the least number of workpieces is the candidate factory.
6. The factory allocation method for distributed job shop scheduling according to claim 5, characterized in that: The priorities for performing the iterative selection include: First-level indicator: maximize the number of new splits of candidate factories; Second-level indicator: Minimize the similarity of new process routes of candidate factories.
7. A factory allocation system for distributed job shop scheduling, characterized in that: include: Data acquisition module: used to obtain workpiece number, process route and processing time; Random sequence generator: used for generating a random artifact sequence based on the artifact number; A division number and similarity calculation unit: used for calculating the division number and process route similarity between workpieces based on the process route; Intelligent allocation module: used to make allocation decisions based on the number of splits and the similarity of the process routes; Load monitoring module: used to track machine utilization and workpiece backlog in each factory in real time.