Task set allocation method for collaborative optimization of production and transportation based on cloud manufacturing

Abstract

The present invention discloses a task set allocation method for collaborative optimization of production and transportation based on cloud manufacturing, which relates to the field of cloud manufacturing production technology. The method includes encoding the allocation of manufacturing service providers of all task sets using a one-dimensional positive integer vector method and randomly generating an initial solution. The present invention realizes efficient modeling of the allocation relationship of task set groups on distributed manufacturing service providers by adopting a one-dimensional positive integer vector encoding method. The method combines the optimal sorting method for minimizing the completion time of a single MSP to simultaneously determine the optimal processing sequence of the workpiece group and the processing sequence of the workpieces within the group. The closed-form expression of the completion time of each machine is obtained based on mathematical derivation, thereby effectively evaluating the performance of the overall scheduling scheme. The method forms a multi-level and multi-angle search collaborative mechanism by introducing a contrast perturbation mechanism, a multi-neighborhood search structure and an iterative greedy strategy, which significantly enhances the diversity retention and deep search capabilities of the algorithm in the solution space.

Description

Technical Field

[0001] The present invention relates to the field of cloud manufacturing production technology, and in particular to a task set allocation method based on collaborative optimization of production and transportation in cloud manufacturing. Background Art

[0002] As the global manufacturing industry accelerates its evolution toward networking, service-oriented, and intelligent manufacturing, the traditional centralized production model faces unprecedented challenges. Increasingly diverse customer demands and shrinking order delivery cycles are driving manufacturers to urgently build highly flexible and responsive production systems. Against this backdrop, Cloud Manufacturing (CM), a next-generation intelligent manufacturing service model that integrates internet-based thinking with advanced manufacturing technologies, enables cross-regional sharing and on-demand access of manufacturing resources through resource virtualization, service-oriented development, and collaboration. Furthermore, CM leverages cluster analysis to divide heterogeneous processing tasks received by the platform into multiple task sets, further improving execution efficiency and significantly enhancing resource allocation efficiency and system responsiveness.

[0003] In a cloud manufacturing environment, the key to improving manufacturing efficiency is to rationally allocate task sets to various Manufacturing Service Providers (MSPs) based on the completed task clustering, scientifically arrange the execution order of task sets on each MSP, and the specific order of tasks within the task set.

[0004] In practice, cloud manufacturing task execution is also influenced by a variety of complex dynamic factors. For example, a set of tasks typically requires a certain amount of preparation time before a specific MSP can begin execution. This time often fluctuates with the task sequence or waiting time, resulting in a time-dependent preparation time. Furthermore, prolonged equipment operation can lead to performance degradation, increasing the processing time of subsequent tasks (also known as the deterioration effect). Furthermore, when continuously processing similar tasks, increasing operator or automation system proficiency can lead to a decrease in unit time (also known as the learning effect).

[0005] In addition, after the task set is processed by its MSP, it needs to be transported directly to the central warehouse by the MSP. Since the distance between each MSP and the warehouse is different, and the unit distance transportation cost of the task set is different, the complexity of logistics cost modeling is further increased.

[0006] The combined effect of the above factors requires that the scheduling model must fully consider the dynamic evolution characteristics of the manufacturing process and the cost diversity of the logistics link in order to more realistically reflect the system operation characteristics and achieve the optimization goals of manufacturing response speed and logistics economy.

[0007] The core idea of ​​Variable Neighborhood Search (VNS) is to systematically alter the neighborhood structure to diversify search directions, thereby avoiding local optimality and enhancing global search capabilities. The algorithm's flexibility provides ample room for future improvements: On the one hand, adaptive adjustment strategies can be designed to enhance the algorithm's adaptability to complex problems; on the other hand, hybrid algorithms can be combined with other optimization strategies to further balance search efficiency and solution quality.

[0008] A collaborative optimization method for the allocation, execution, and transportation of cloud manufacturing task sets has been proposed in traditional optimization strategies, taking into account a variety of complex dynamic factors. The aim is to achieve efficient allocation of manufacturing resources and reasonable coordination of logistics and transportation in a distributed manufacturing environment. This method targets the task sets divided by the cloud manufacturing platform, comprehensively considering the processing capabilities, geographical location, and load status of each manufacturing service provider (MSP) to achieve a reasonable allocation of task sets. To solve this problem efficiently, this method constructs a hybrid optimization framework based on the improved variable neighborhood search (IVNS) and iterated greedy (IG) strategies, achieving collaborative scheduling optimization of manufacturing and logistics links with the goal of minimizing the weighted maximum completion time and distribution cost. The basic process of the above algorithm includes the following steps:

[0009] ① Initialize algorithm parameters and initial solution;

[0010] ② Perturb the current solution to escape the local optimal trap;

[0011] ③ Apply the neighborhood structure to perform a local search for the solution. If a better solution is found, return to step ② to continue iterating; otherwise, switch to the next neighborhood structure;

[0012] ④ Determine whether the termination condition of the variable neighborhood search is met. If so, end the search at this stage;

[0013] ⑤ Enable the iterative greedy strategy to further optimize the current solution;

[0014] ⑥ Check the termination condition of the entire algorithm. If it is met, the operation ends; otherwise, return to step ② and continue a new round of search.

[0015] Traditional technologies have the following shortcomings:

[0016] 1. Research issues

[0017] There are still many key issues in the current task scheduling methods in cloud manufacturing environments that have not been effectively resolved, especially in the dynamic effect modeling of task execution time and the coordinated optimization of production efficiency and transportation costs.

[0018] First, most methods still use a single task as the basic unit of scheduling, lacking an understanding of the characteristics of task sets. In practical applications, clustering tasks with similar processes into task sets through grouping techniques for unified scheduling can not only reduce equipment switching frequency and setup time, but also significantly improve resource utilization efficiency. However, existing methods often overlook the synergistic advantages brought by task set clustering, making it difficult to fully realize its potential in mass customization production models.

[0019] Secondly, during the modeling process, most methods fail to fully consider the dynamic factors common in the manufacturing process. For example, a set of tasks typically requires a certain amount of preparation time before a manufacturing service provider can begin executing them. This time often varies with the scheduling order or waiting time, exhibiting a significant time-dependent characteristic. However, most current models still treat this as a fixed value, making it difficult for scheduling plans to accurately reflect the actual production status, affecting the feasibility and stability of the overall scheduling scheme. Furthermore, while some methods have addressed the deterioration effect during task execution (such as increased task processing time due to equipment aging or operator fatigue), few have also considered the learning effect (i.e., efficiency gains resulting from increased operator proficiency as similar tasks are repeatedly executed). The combined impact of these two factors on task completion time cannot be ignored in real systems. Ignoring their modeling will cause scheduling strategies to deviate from actual operating rules, reducing the robustness and adaptability of scheduling results.

[0020] Finally, existing cloud manufacturing scheduling methods generally lack effective integration of logistics and transportation, failing to fully consider the need for immediate transportation after task completion. Because different task sets have significantly different characteristics, such as transportation costs, optimizing scheduling solely from the manufacturing side without integrating logistics factors will make it difficult to achieve system-level optimization.

[0021] (2) Research methods

[0022] From the perspective of algorithm design, current cloud manufacturing task scheduling methods generally have the following limitations when considering complex features such as task set characteristics, time-dependent preparation time, deterioration-learning effects, and logistics and transportation coupling:

[0023] On the one hand, most scheduling optimization methods lack precise modeling of task completion times. This is especially true when multiple dynamic factors are involved. They fail to derive a completion time expression that reflects the combined effects of time-dependent setup time, degradation effects, and learning effects. This makes it difficult for algorithms to accurately evaluate the actual performance of different scheduling schemes, limiting their application in complex manufacturing environments.

[0024] On the other hand, traditional algorithms also have limitations in their search mechanism design. Many methods employ only a single neighborhood structure for local search, lacking effective perturbation mechanisms and the ability to explore diverse solutions. This makes them prone to falling into local optima, resulting in slow convergence and low-quality solutions. This is particularly true when dealing with high-dimensional combinatorial optimization problems, such as those involving multiple task sets and multiple MSPs. The algorithms' global search capabilities and stability are significantly insufficient.

[0025] Therefore, there is an urgent need for a new method that integrates task set-oriented manufacturing and logistics collaborative scheduling modeling and an efficient hybrid optimization algorithm, which can achieve efficient solution of task scheduling in a cloud manufacturing environment while taking into account both manufacturing responsiveness and logistics initiative.

[0026] In view of the above technical defects, a solution is now proposed. Summary of the Invention

[0027] This paper aims to address the collaborative optimization problem of task allocation, execution, and transportation for multiple manufacturing service providers (MSPs) in a cloud manufacturing environment. This problem involves the coupling of task set characteristics, time-dependent setup times, the deterioration-learning effect, and just-in-time transportation mechanisms. By proposing a two-stage hybrid optimization approach, this approach achieves efficient collaborative scheduling of manufacturing and logistics while balancing maximum makespan and transportation costs, improving overall system responsiveness and resource utilization.

[0028] (1) Determination of the sequence of single MSP task sets and internal tasks

[0029] When each MSP has obtained the information of the assigned task set, a scheduling model and optimal scheduling method that considers dynamic factors are built within it:

[0030] First, the efficiency decline caused by equipment wear and operator fatigue (deterioration effect) and the improvement in operator proficiency brought about by repeatedly performing similar tasks (learning effect) are quantified, and a dynamic processing time expression is established. Second, a time-dependent preparation time model is introduced, in which the preparation time before the start of a task set fluctuates with its scheduling order and waiting time in the MSP. Finally, the optimal sorting strategy for minimizing the completion time of a single MSP is obtained through mathematical derivation, and the processing sequence of the task set and its internal tasks on each MSP is determined to achieve the optimization of local scheduling performance.

[0031] (2) Multi-MSP Task Set Allocation and Logistics Collaborative Optimization

[0032] In the second phase, a hybrid optimization framework that integrates the improved variable neighborhood search (IVNS) and iterative greedy (IG) strategy is used to achieve global optimal allocation of task sets among multiple MSPs and coordinate the transportation costs of each task set. Specifically:

[0033] In the IVNS phase, various neighborhood structures are designed to enhance the ability to explore diverse solution spaces. A contrast perturbation mechanism is introduced to prevent the algorithm from falling into local optima and improve convergence stability. In the IG phase, system load balancing is achieved by destroying the task sets of some high-loaded MSPs and rebuilding them on low-loaded nodes. Furthermore, a heuristic decoding mechanism is combined with an optimal sorting strategy driven by mathematical properties to further improve the computational efficiency and solution quality of the scheduling scheme. Furthermore, to enhance final delivery efficiency, this method also introduces an immediate post-task transportation mechanism. This mechanism means that after each task in its corresponding task set is completed, the corresponding MSP is directly transported to its designated destination warehouse. The total transportation cost is calculated based on the transportation distance and the unit distance transportation cost of the task set, and is incorporated into the scheduling objective function for unified optimization.

[0034] In summary, the present invention can provide high-quality scheduling decisions for cloud manufacturing platforms, effectively improving the comprehensive effects of manufacturing system response speed and logistics economy, and is particularly suitable for intelligent manufacturing application scenarios with intensive personalized orders and high coordination between manufacturing and logistics.

[0035] To achieve the above objectives, the present invention adopts the following technical solution: a task set allocation method based on collaborative optimization of production and transportation in cloud manufacturing, comprising the following steps:

[0036] Step 1: Use a one-dimensional positive integer vector to encode the allocation of manufacturing service providers for all task sets and randomly generate an initial solution. ;

[0037] Step 2: Set the maximum number of iterations of the optimized task set allocation algorithm , initialize the number of iterations: , and set the neighborhood structure number: ;

[0038] Step 3: Calculate the initial solution Fitness , and list the historical optimal solutions Initialized to an empty list;

[0039] Step 4: If the current neighborhood structure number , then jump to step five, otherwise jump to step seven;

[0040] Step 5: Current solution Perform the contrast perturbation operation to obtain the perturbation solution , and then Using Neighborhood Operations Conduct local search to obtain new solutions , and calculate its fitness ;

[0041] Step 6: If , that is, the new solution is the current solution, then update the current solution , and reset the neighborhood numbers: ; otherwise update the neighbor number: , return to step 4;

[0042] Step 7: Current solution Execute iterative greedy strategy to generate improved solutions , and add it to the historical optimal solution list middle;

[0043] Step 8. Update the number of iterations: ,if , then return to step 4, otherwise go to step 9;

[0044] Step 9: Traverse the historical optimal solution list , find the optimal individual with the smallest fitness , as the final output of the algorithm.

[0045] Furthermore, the encoding process is: using a one-dimensional vector using a one-dimensional vector Encode the solution, where the vector elements The value range is 1 to Any integer in .

[0046] Furthermore, the decoding process is: The value of represents the task set The corresponding manufacturing service provider number assigned is decoded according to this rule , obtain all task sets on each manufacturing service provider; use the optimal sorting that minimizes the completion time of a single manufacturing service provider for each manufacturing service provider to obtain the corresponding optimal task set execution order and the task execution order within the group.

[0047] Furthermore, the fitness calculation process is as follows: the fitness of the solution vector is represented by the weighted sum of the maximum completion time and the total transportation cost after the optimal sorting to minimize the completion time of a single manufacturing service provider. According to mathematical derivation, the completion time formula of each manufacturing service provider can be obtained: ,in , , accordingly, the solution Fitness , the smaller the value, the better the corresponding solution, where, is the completion time of manufacturing service provider numbered i, is the position number of the task set, that is, the gth task set to be processed, is the total number of task sets, Is the manufacturing service provider numbered i The dynamic characteristic factors of the task set being processed, Is the manufacturing service provider numbered i The preparation time deterioration effect factor of the task set being processed, Is the manufacturing service provider numbered i The execution time characteristic factor of the processed task set, Is the manufacturing service provider numbered i The normal preparation time for a set of tasks to be processed, Is the manufacturing service provider numbered i The number of tasks in the processed task set, Is the manufacturing service provider numbered i The execution time learning factor of the processed task set, is the manufacturing service provider numbered i The execution time deterioration factor of the processed task set, is the manufacturing service provider numbered i The normal execution time of the rth task in the processed task set, Is the manufacturing service provider numbered i The dynamic characteristic factor of the rth executed task in the set of processed tasks, is the weighted coefficient of the maximum completion time, is the weighted coefficient of logistics cost, is the total number of manufacturing service providers, is a binary variable, is the distance from the manufacturing service provider numbered i to the central warehouse, yes The transportation cost per unit distance of the task set.

[0048] Furthermore, the optimal sorting method to minimize the completion time of a single manufacturing service provider is obtained through complete mathematical derivation. This algorithm is used to arrange the execution order of the task sets on the manufacturing service provider and the execution order of the workpieces within the group.

[0049] The specific steps of the optimal algorithm are as follows:

[0050] S11: For each task set on the MSP, the tasks in the group are executed according to the normal execution time. Arrange in non-decreasing order;

[0051] S12. For each workpiece group on the manufacturing service provider, calculate the corresponding task set weight after the above sorting. ,in , ,in, is the task set number, It is numbered The dynamic characteristic factor of the task set, It is numbered The preparation time deterioration effect factor of the task set, It is numbered The execution time characteristic factor of the task set, It is numbered The normal preparation time of the task set, It is numbered The number of tasks in the task set, It is numbered The execution time learning factor of the task set is It is numbered The execution time degradation factor of the task set is It is numbered The normal execution time of the rth task in the task set, It is numbered The dynamic characteristic factor of the rth executed task in the task set;

[0052] S13, the workpiece groups on the workpiece manufacturing machines of the manufacturing service provider are arranged according to their weights Arrange in non-decreasing order and execute.

[0053] Furthermore, the specific steps of the contrast perturbation mechanism are as follows:

[0054] S21. For the solutions entering the perturbation process, apply the preset perturbation operations respectively. 、 ;

[0055] S22, compare the fitness of the two new solutions obtained with the original solution;

[0056] S23. Select the best solution among the three as the output of the disturbance phase.

[0057] Furthermore, the constructed neighborhood structure is as follows: Neighborhood structure Neighborhood structure : Randomly select an element and an insertion position in the vector, and insert the element before the selected insertion position; neighborhood structure : Randomly select a continuous sub-segment of any length in the vector and reverse the order of all elements in the sub-segment, while keeping the rest of the vector unchanged; neighborhood structure : Randomly select two non-overlapping contiguous subsequences in a vector and swap the positions of the two subsequences, while keeping the order of the remaining elements unchanged.

[0058] Furthermore, the iterative greedy strategy is as follows: the manufacturing service provider with the longest current completion time is identified, the task set with the smallest weight is selected from it, and the task set is transferred to the manufacturing service provider with the shortest current completion time for execution.

[0059] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0060] 1. This paper addresses the complex issues of task set allocation, execution, and transportation coordination in cloud manufacturing environments by proposing a hybrid optimization algorithm that integrates an improved variable neighborhood search (IVNS) with an iterative greedy (IG) strategy. By employing a one-dimensional positive integer vector encoding scheme, this algorithm efficiently models the allocation relationships between task sets across distributed manufacturing service providers (MSPs). In combination with an optimal sorting method that minimizes the completion time of individual MSPs, it simultaneously determines the optimal processing sequence for both groups of workpieces and within groups. Furthermore, a closed-form expression for the completion time of each machine is derived mathematically, effectively evaluating the performance of the overall scheduling solution.

[0061] 2. The present invention focuses on dynamic factors in actual manufacturing scenarios, including the deterioration effect caused by equipment wear and personnel fatigue, the learning effect brought about by improved operating proficiency, and time-dependent preparation time, etc., and achieves precise optimization through systematic innovative methods. Specifically, a completion time calculation model that incorporates deterioration and learning factors is constructed to convert the volatility in the task execution process into quantifiable indicators; at the same time, a time-dependent preparation time modeling method that conforms to actual production laws is proposed to accurately depict the dynamic characteristics of preparation time over time. The organic combination of the above technical solutions has significantly improved the applicability and robustness of the present invention in real production environments.

[0062] 3. By introducing a contrast perturbation mechanism, a multi-neighborhood search structure, and an iterative greedy strategy, this invention forms a multi-level, multi-angle search coordination mechanism, significantly enhancing the algorithm's ability to maintain diversity and deep search within the solution space. This mechanism not only effectively prevents the algorithm from falling into local optima but also demonstrates greater adaptability and flexibility in handling scheduling problems involving task allocation and transport coordination. Through these innovative designs, this invention demonstrates superior performance in improving manufacturing response speed and reducing logistics costs, possessing broad application prospects and promotional value. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A schematic flow chart of the method of the present invention is shown. DETAILED DESCRIPTION

[0064] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention. Example

[0065] like Figure 1 As shown in the figure, the present invention optimizes the task set scheduling problem in a cloud manufacturing environment by designing a hybrid metaheuristic algorithm based on improved variable neighborhood search (IVNS) and iterative greedy (IG) strategy, considering the task set characteristics, time-dependent preparation time, deterioration-learning combined effect and instant transportation mechanism.

[0066] (1) Example description

[0067] Cloud manufacturing platform will be the original Heterogeneous processing tasks are clustered into an indivisible set of tasks, assigned to Processing is performed on distributed manufacturing service providers (MSPs). Each MSP must rationally arrange the execution order of task sets and tasks within them. When an MSP switches to a new task set for execution, a certain amount of preparation time is required; however, tasks within the same task set can execute continuously without additional preparation time. Each MSP can only process one task set at a time. All tasks are available for production at time zero, and all MSPs are initially idle. Once a task begins execution, it cannot be interrupted, and unwarranted waiting is not permitted.

[0068] Considering the differences in processing capabilities of distributed MSPs, the execution time of the same task in different MSPs is different. The actual execution time of the task is affected by the combined effects of equipment performance degradation (deterioration effect) and operator proficiency improvement (learning effect). MSP processing Task Focus The actual execution time of a task can be expressed as: .in: The standard execution time of the task in the MSP; is the learning factor of the corresponding task set; is the deterioration factor of the corresponding task set; It is the cumulative running time of the MSP from time zero to the start of the task execution.

[0069] At the same time, the preparation time required for task set switching is affected by the current MSP operation status and presents dynamic fluctuation characteristics. The MSP switches to The actual preparation time for each task set is: .in: is the standard preparation time for the task set, The degradation factor for the preparation time of the task set, It is the cumulative running time of the MSP from time zero to switching to the task set.

[0070] After completion, all task sets will be transported directly to the central warehouse by the MSPs they perform. The distance from each MSP to the central warehouse is different, and the unit distance transportation cost of each task set is also different.

[0071] The optimization goal of this problem is to achieve the coordinated optimization of manufacturing response speed and logistics transportation costs, so that the maximum completion time of each MSP and the weighted sum of the total transportation costs of all task sets are minimized: .in: is the maximum completion time of each MSP, It corresponds to the distance from MSP to the central warehouse. is the transportation cost per unit distance of the corresponding task set, and is the weight coefficient of the corresponding sub-goal, is a binary variable (only when the task set Assigned to number The value is 1 when it is MSP, and 0 otherwise).

[0072] (2) Optimization method

[0073] The specific steps of the optimization method proposed in the present invention are as follows:

[0074] Step 1: Use a one-dimensional positive integer vector to encode the MSP allocation of all task sets and randomly generate an initial solution. , where the encoding process is: using a one-dimensional vector Encode the solution, where the vector elements The value range is 1 to Any integer in .

[0075] The decoding process is: The value of represents the task set The corresponding MSP number assigned. Decode according to this rule , we can obtain all the task sets on each MSP; then, for each MSP, we use the optimal sorting method that minimizes the completion time of a single MSP to obtain the corresponding optimal task set execution order and the task execution order within the group. For the optimal sorting method that minimizes the completion time of a single MSP, through complete mathematical derivation, we derive the optimal algorithm that minimizes its completion time for a single MSP. This algorithm is used to arrange the execution order of the task sets on that MSP and the execution order of the workpieces within the group.

[0076] The specific steps of the optimal algorithm are as follows:

[0077] S11: For each task set on the MSP, the tasks in the group are executed according to the normal execution time. Arrange in non-decreasing order;

[0078] S12: For each workpiece group on the MSP, calculate the corresponding task set weight after the above sorting. ,in , , is the number of the task set, that is, the name of the task set, such as task set 1, task set 2, task set 3, It is numbered The dynamic characteristic factor of the task set, It is numbered The preparation time deterioration effect factor of the task set, It is numbered The execution time characteristic factor of the task set, It is numbered The normal preparation time of the task set, It is numbered The number of tasks in the task set, It is numbered The execution time learning factor of the task set is It is numbered The execution time degradation factor of the task set is It is numbered The normal execution time of the rth task in the task set, It is numbered The dynamic characteristic factor of the rth executed task in the task set;

[0079] S13, group the workpieces on the machine according to their weights Arrange in non-decreasing order and execute.

[0080] Step 2: Set the maximum number of iterations of the algorithm , initialize the number of iterations: , and set the neighborhood structure number: The neighborhood structure determines the exploration capability and convergence efficiency of Variable Neighborhood Search (VNS). Reasonable and effective neighborhood design can effectively avoid falling into local optimality. To this end, we constructed the following three different neighborhood structures:

[0081] 1. Neighborhood structure (Element Insertion): Randomly selects an element and an insertion position in the vector, and inserts the element before the selected insertion position.

[0082] 2. Neighborhood Structure (Fragment Reversal): Randomly select a continuous sub-segment of any length in a vector and reverse the order of all elements in the sub-segment, while keeping the rest of the vector unchanged.

[0083] 3. Neighborhood structure (Fragment Swap): Randomly select two non-overlapping contiguous subsequences in a vector and swap the positions of the two subsequences, while keeping the order of the remaining elements unchanged.

[0084] Step 3: Calculate the initial solution Fitness , and list the historical optimal solutions Initialized to an empty list; the fitness calculation process is as follows: the fitness of the solution vector is represented by the weighted sum of the maximum completion time and the total transportation cost after arrangement according to the above requirements. According to mathematical derivation, the completion time formula of each manufacturing service provider can be obtained: ,in , , accordingly, the solution Fitness , the smaller the value, the better the corresponding solution, where, is the completion time of manufacturing service provider numbered i, is the position number of the task set, that is, the gth task set to be processed, is the total number of task sets, Is the manufacturing service provider numbered i The dynamic characteristic factors of the task set being processed, Is the manufacturing service provider numbered i The preparation time deterioration effect factor of the task set being processed, Is the manufacturing service provider numbered i The execution time characteristic factor of the processed task set, Is the manufacturing service provider numbered i The normal preparation time for a set of tasks to be processed, Is the manufacturing service provider numbered i The number of tasks in the processed task set, Is the manufacturing service provider numbered i The execution time learning factor of the processed task set, is the manufacturing service provider numbered i The execution time deterioration factor of the processed task set, is the manufacturing service provider numbered i The normal execution time of the rth task in the processed task set, Is the manufacturing service provider numbered i The dynamic characteristic factor of the rth executed task in the set of processed tasks, is the weighted coefficient of the maximum completion time, is the weighted coefficient of logistics cost, is the total number of manufacturing service providers, is a binary variable (only if The task set is assigned to the manufacturing service provider with the value of 1, and 0 otherwise). is the distance from the manufacturing service provider numbered i to the central warehouse, yes The transportation cost per unit distance of the task set.

[0085] Step 4: If the current neighborhood structure number , then jump to step five, otherwise jump to step seven;

[0086] Step 5: Current solution Perform the contrast perturbation operation to obtain the perturbation solution , and then Using Neighborhood Operations Conduct local search to obtain new solutions , and calculate its fitness The contrastive shaking mechanism in this paper is different from the perturbation operation in traditional variable neighborhood search (VNS). This paper introduces a contrastive shaking mechanism (CS) to improve the diversity of the solution space and reduce invalid searches. Here, we first explain the two perturbation structures used in this operation:

[0087] 1. Disturbance structure (Fragment Reset): Randomly select a continuous sub-segment of arbitrary length in the vector, and randomly reassign the elements in the sub-segment within the value range of the solution elements, while the rest of the vector remains unchanged.

[0088] 2. Disturbance structure (Fragment Insertion): Randomly selects a continuous subsegment and an insertion position outside the subsegment, and then inserts the selected subsegment before the insertion position.

[0089] The specific steps of the contrast perturbation mechanism are as follows:

[0090] S21. For the solutions entering the perturbation process, apply the above perturbation operations respectively. 、 ;

[0091] S22, compare the fitness of the two new solutions obtained with the original solution;

[0092] S23. Select the best solution among the three as the output of the disturbance phase.

[0093] Step 6: If , that is, the new solution is the current solution, then update the current solution , and reset the neighborhood numbers: ; otherwise update the neighbor number: , return to step 4;

[0094] Step 7: Current solution Execute iterative greedy strategy to generate improved solutions , and add it to the historical optimal solution list To significantly improve the algorithm's convergence efficiency and deepen the exploration of the solution space, this paper proposes an innovative iterative greedy (IG) strategy. In its implementation, the algorithm identifies the MSP with the longest current completion time, selects the set of tasks with the lowest weight, and transfers them to the MSP with the shortest current completion time. This process is repeated until the solution is no longer optimized. This dynamic allocation adjustment mechanism, through iterative task migration, continuously optimizes the load distribution across MSPs, promoting balanced utilization of system resources and effectively reducing the overall maximum completion time.

[0095] To ensure the algorithm's robust global search capabilities and avoid being trapped in local optimal solutions, this strategy specifically designs the scope of its iterative greedy operations. Specifically, this strategy optimizes only the solution generated by each iteration of the algorithm. The results of these operations are not recirculated in other stages. Instead, the optimized solutions are collected and temporarily stored in a dedicated list. After all iterations are complete, the algorithm compares the solutions in the list and selects the optimal solution as the final output. This approach not only ensures that the algorithm can fully explore the diversity of the solution space during iteration, but also ensures global optimality in the final output through a centralized screening mechanism, achieving an effective balance between the algorithm's convergence speed and solution quality.

[0096] Step 8. Update the number of iterations: .if , then return to step 4, otherwise go to step 9;

[0097] Step 9: Traverse the historical optimal solution list , find the optimal individual with the smallest fitness , as the final output of the algorithm.

[0098] The size of the interval and threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by technical personnel in this field for each set of sample data; as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0099] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by those skilled in the art according to actual conditions.

[0100] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A task set allocation method for collaborative optimization of production and transportation based on cloud manufacturing, characterized by: The following steps are involved: Step 1: Use a one-dimensional positive integer vector to encode the allocation of manufacturing service providers for all task sets and randomly generate an initial solution. ; Step 2: Set the maximum number of iterations of the optimized task set allocation algorithm , initialize the number of iterations: , and set the neighborhood structure number: ; Step 3: Calculate the initial solution Fitness , and list the historical optimal solutions Initialized to an empty list; Step 4: If the current neighborhood structure number , then jump to step five, otherwise jump to step seven; Step 5: Current solution Perform the contrast perturbation operation to obtain the perturbation solution , and then Using Neighborhood Operations Conduct local search to obtain new solutions , and calculate its fitness ; Step 6: If , that is, the new solution is the current solution, then update the current solution , and reset the neighborhood numbers: ; otherwise update the neighbor number: , return to step 4; Step 7: Current solution Execute iterative greedy strategy to generate improved solutions , and add it to the historical optimal solution list middle; Step 8. Update the number of iterations: ,if , then return to step 4, otherwise go to step 9; Step 9: Traverse the historical optimal solution list , find the optimal individual with the smallest fitness , as the final output of the algorithm.

2. The method for allocating task sets for collaborative optimization of production and transportation based on cloud manufacturing according to claim 1 is characterized in that: The encoding process is: using a one-dimensional vector Encode the solution, where the vector elements The value range is 1 to Any integer in .

3. The method for allocating task sets for collaborative optimization of production and transportation based on cloud manufacturing according to claim 1 is characterized in that: The decoding process is: The value of represents the task set The corresponding manufacturing service provider number assigned is decoded according to this rule , obtain all task sets on each manufacturing service provider; use the optimal sorting that minimizes the completion time of a single manufacturing service provider for each manufacturing service provider to obtain the corresponding optimal task set execution order and the task execution order within the group.

4. The method for allocating task sets for collaborative optimization of production and transportation based on cloud manufacturing according to claim 1 is characterized in that: The fitness calculation process is as follows: the fitness of the solution vector is represented by the weighted sum of the maximum completion time and the total transportation cost after the optimal sorting to minimize the completion time of a single manufacturing service provider. According to mathematical derivation, the completion time formula of each manufacturing service provider can be obtained: ,in , , accordingly, the solution Fitness , the smaller the value, the better the corresponding solution, where, is the completion time of manufacturing service provider numbered i, is the position number of the task set, that is, the gth task set to be processed, is the total number of task sets, Is the manufacturing service provider numbered i The dynamic characteristic factors of the task set being processed, Is the manufacturing service provider numbered i The preparation time deterioration effect factor of the task set being processed, Is the manufacturing service provider numbered i The execution time characteristic factor of the processed task set, Is the manufacturing service provider numbered i The normal preparation time for a set of tasks to be processed, Is the manufacturing service provider numbered i The number of tasks in the processed task set, Is the manufacturing service provider numbered i The execution time learning factor of the processed task set, is the manufacturing service provider numbered i The execution time deterioration factor of the processed task set, is the manufacturing service provider numbered i The normal execution time of the rth task in the processed task set, Is the manufacturing service provider numbered i The dynamic characteristic factor of the rth executed task in the set of processed tasks, is the weighted coefficient of the maximum completion time, is the weighted coefficient of logistics cost, is the total number of manufacturing service providers, is a binary variable, is the distance from the manufacturing service provider numbered i to the central warehouse, yes The transportation cost per unit distance of the task set.

5. The method for allocating task sets for collaborative optimization of production and transportation based on cloud manufacturing according to claim 3 is characterized in that: The optimal sorting method to minimize the completion time of a single manufacturing service provider is developed through complete mathematical derivation. The optimal algorithm to minimize the completion time of a single manufacturing service provider is used to arrange the execution order of the task set on the manufacturing service provider and the execution order of the workpieces within the group. The specific steps of the optimal algorithm are as follows: S11: For each task set on the manufacturing service provider, the tasks in the group are executed according to the normal execution time. Arrange in non-decreasing order; S12. For each workpiece group on the manufacturing service provider, calculate the corresponding task set weight after the above sorting. ,in , ,in, is the task set number, It is numbered The dynamic characteristic factor of the task set, It is numbered The preparation time deterioration effect factor of the task set, It is numbered The execution time characteristic factor of the task set, It is numbered The normal preparation time of the task set, It is numbered The number of tasks in the task set, It is numbered The execution time learning factor of the task set is It is numbered The execution time degradation factor of the task set is It is numbered The normal execution time of the rth task in the task set, It is numbered The dynamic characteristic factor of the rth executed task in the task set; S13, the workpiece groups on the workpiece manufacturing machines of the manufacturing service provider are arranged according to their weights Arrange in non-decreasing order and execute.

6. The method for allocating task sets for collaborative optimization of production and transportation based on cloud manufacturing according to claim 1 is characterized in that: The specific steps of the contrast perturbation mechanism are as follows: S21. For the solutions entering the perturbation process, apply the preset perturbation operations respectively. 、 ; S22, compare the fitness of the two new solutions obtained with the original solution; S23. Select the best solution among the three as the output of the disturbance phase.

7. The method for allocating task sets for collaborative optimization of production and transportation based on cloud manufacturing according to claim 1 is characterized in that: The constructed neighborhood structure is as follows: Neighborhood structure : Randomly select an element and an insertion position in the vector, and insert the element before the selected insertion position; neighborhood structure : Randomly select a continuous sub-segment of any length in the vector and reverse the order of all elements in the sub-segment, while keeping the rest of the vector unchanged; neighborhood structure : Randomly select two non-overlapping contiguous subsequences in a vector and swap the positions of the two subsequences, while keeping the order of the remaining elements unchanged.

8. The method for allocating task sets for collaborative optimization of production and transportation based on cloud manufacturing according to claim 1 is characterized in that: The iterative greedy strategy is as follows: identify the manufacturing service provider with the longest current completion time, select the task set with the smallest weight, and transfer it to the manufacturing service provider with the shortest current completion time for execution.

Images