Flow shop group scheduling multi-objective optimization method and system of dual-evolution mechanism
Through the multi-objective optimization method of flow workshop group scheduling with dual evolution mechanism, the heterogeneous graph model and multiple group co-evolution algorithm are used to solve the problems of dynamic events and multi-objective conflicts in flexible work workshop scheduling, and a high-quality scheduling solution is generated, which improves the diversity and robustness of scheduling.
Patent Information
- Application Number
- CN202510476694.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-16
- Publication Date
- 2025-07-18
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing technology is difficult to deal with real-time dynamic events in flexible operation workshop scheduling, and the single-objective optimization model cannot coordinate the conflict between multiple targets. The application of traditional graph neural networks in scheduling is limited to isomorphic graph modeling, and the process-machine heterogeneous relationship is not fully mined, resulting in poor robustness of the scheduling scheme and insufficient feature expression capabilities.
The multi-objective optimization method of flow workshop group scheduling using a dual evolution mechanism simplifies the correlation expression of the process and the machine through a heterogeneous graph model, combines multiple group co-evolution algorithms to generate high-quality training data, and uses a progressive attention network and a near-end strategy optimization algorithm to achieve iterative optimization of dynamic scheduling strategies.
Effectively solve the problems of processing division and speed allocation of group sequences and workpiece sequences, generate high-quality initial populations, enhance the convergence of population diversity and reconciliation, optimize target values, and improve the diversity and robustness of scheduling schemes.
Smart Images
Figure CN120335406A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of scheduling analysis and processing methods for flexible job shops, and particularly to a multi-objective optimization method and system for flow shop group scheduling with a dual evolution mechanism. Background Art
[0002] The Flexible Job-Shop Scheduling Problem (FJSP) is a core optimization problem in the manufacturing field. Its goal is to reasonably allocate processing tasks to minimize the makespan, improve resource utilization, or reduce delay costs while satisfying the process sequence constraints and machine resource limitations. Flexible job shop scheduling (FJSP) is a core issue in intelligent manufacturing, and it is necessary to balance dynamic objectives (such as minimizing tardiness, balancing machine loads, and maximizing process completion rates) in a complex environment with multiple processes, multiple machines, and multiple constraints.
[0003] Traditional methods mainly rely on heuristic rules (such as Shortest Processing Time First - SPT, Earliest Due Date First - EDD) and meta - heuristic algorithms, but these methods have significant defects:
[0004] Firstly, scheduling strategies based on static assumptions are difficult to cope with real - time dynamic events (such as machine failures, emergency order insertions), resulting in poor robustness of the solutions;
[0005] Secondly, single - objective optimization models cannot coordinate and handle the conflicts between multiple objectives. For example, shortening the due date may lead to uneven machine loads, and existing research mostly focuses on single objectives, making it difficult to meet the actual production requirements;
[0006] Thirdly, the application of traditional Graph Neural Networks (GNNs) in scheduling is limited to homogeneous graph modeling (only considering a single node type of process or machine), and does not fully explore the heterogeneous relationships between processes and machines (such as the selection dependence of processes on machines, the difference in cooperation capabilities between machines) and the sequential constraints between processes (such as the strict front - back relationship of internal processes of workpieces), resulting in insufficient feature expression capabilities. Summary of the Invention
[0007] In an exemplary embodiment of the present application, a multi - objective optimization method and system for flow shop group scheduling with a dual evolution mechanism are provided. By using a heterogeneous graph model to simplify the association expression between processes and machines, combined with a multi - population cooperative evolution algorithm to generate high - quality training data, and using a progressive attention network and a proximal policy optimization algorithm to achieve iterative optimization of dynamic scheduling strategies.
[0008] On the one hand, the present application provides a multi - objective optimization method for flow shop group scheduling with a dual evolution mechanism, which includes the following steps:
[0009] S1: Generate an initial solution set using a hybrid method that includes a random partitioning strategy and a heuristic strategy. The heuristic strategy includes a group partitioning strategy GSO, a group sequence initialization strategy GSST, and a workpiece sequence initialization strategy PF.
[0010] S2: Divide the initial population into a non-dominated solution set P_A and a dominated solution set P_B according to the quality of the solutions.
[0011] S3: Before each iteration, through the pre-evolution evaluation of the temporary population P_tem, select a cross-evolution search CES or an improved particle swarm optimization IPSO as the evolution strategy for the non-dominated solution set P_A and the dominated solution set P_B respectively.
[0012] S4: Execute the selected evolution strategies for the non-dominated solution set P_A and the dominated solution set P_B respectively, and ensure balance by adjusting the population size.
[0013] S5: Iteratively update the solution set until the termination condition is met, and output the Pareto optimal set of non-dominated solutions.
[0014] Furthermore, in step S1, the encoding scheme of the hybrid initialization includes:
[0015] Divide the group sequence G into an internal processing group G_in and an outsourcing processing group G_out;
[0016] Divide the workpiece sequence π into an internal processing workpiece π_in and an outsourcing workpiece π_out;
[0017] Assign a speed level v to each internal processing workpiece, and the speed level corresponds to a preset set of machining speed values.
[0018] Furthermore, the group partitioning strategy GSO includes:
[0019] Arrange in ascending order of the total processing time of the workpieces within the group;
[0020] Based on the optimization objectives of cost and production efficiency, determine the optimal partitioning of the internal processing group.
[0021] Furthermore, the group sequence initialization strategy GSST includes:
[0022] Construct a matrix of the total setup time between groups;
[0023] Select the group sequence with the minimum setup time to generate the initial sorting of the internal processing group.
[0024] Furthermore, the workpiece sequence initialization strategy PF includes:
[0025] Generate the workpiece processing order by the dynamic insertion method according to the principle of minimizing the total blocking and idle time of the workpieces within the group.
[0026] Furthermore, the evaluation method for the dynamic policy selection in step S3 is as follows:
[0027] Calculate the fitness value of the solution based on the Penalty Boundary Intersection method PBI, and the formula is:
[0028] f PBI = R1 + θR2.
[0029] Furthermore, the Cross Evolution Search CES includes:
[0030] Random insertion operation of workpieces within a group;
[0031] Group sequence optimization based on block move search;
[0032] Position-based crossover PBC operation, select parent solutions through roulette wheel selection and generate offspring sequences.
[0033] On the other hand, the present application also provides a flow shop group scheduling multi-objective optimization system with a dual evolution mechanism, which is used to execute the method provided in the first aspect above, and it includes:
[0034] Initialization module, which is configured to generate an initial solution set by using a hybrid method including a random partitioning strategy and a heuristic strategy, and the heuristic strategy includes a group partitioning strategy GSO, a group sequence initialization strategy GSST, and a workpiece sequence initialization strategy PF;
[0035] Partitioning module, which is configured to partition the initial population into a non-dominated solution set P_A and a dominated solution set P_B according to the quality of the solutions;
[0036] Strategy determination module, which is configured to, before each iteration, select the Cross Evolution Search CES or the Improved Particle Swarm Optimization IPSO as the evolution strategy for the non-dominated solution set P_A and the dominated solution set P_B respectively through the pre-evolution evaluation of the temporary population P_tem;
[0037] Evolution module, which is configured to execute the selected evolution strategies on the non-dominated solution set P_A and the dominated solution set P_B respectively, and ensure balance by adjusting the population size;
[0038] Pareto set generation module, which is configured to iteratively update the solution set until the termination condition is met, and output the Pareto optimal set of non-dominated solutions.
[0039] The embodiments of the present application have the following beneficial effects: the multi-objective optimization method and system for flow shop group scheduling with dual evolution mechanism of the present invention, the multi-objective evolutionary algorithm of the dual evolution mechanism is used to solve specific group sequence and workpiece sequence processing optimization problems, in the encoding and decoding links, the unique encoding scheme effectively solves the processing division and speed distribution problems of group sequence and workpiece sequence, decoding can obtain feasible solutions and calculate target values based on relevant information, and through three heuristic methods and random partitioning strategies, it fully utilizes problem-specific knowledge and enhances population diversity to generate high-quality initial populations. It is divided into two populations according to the dominance relationship of the solution, and the penalty-based boundary crossing method is used to select the appropriate evolutionary strategy to ensure the convergence and diversity of the solution. The algorithm can effectively solve related problems and performs well in optimizing target values, improving the quality and diversity of solutions, etc. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0041] Figure 1 A schematic diagram of a multi-objective optimization method for flow shop group scheduling with a dual evolution mechanism provided in an embodiment of the present application is exemplarily shown;
[0042] Figure 2 An algorithm framework diagram of a multi-objective optimization method for flow shop group scheduling with a dual evolution mechanism provided in an embodiment of the present application is exemplarily shown;
[0043] Figure 3 A schematic diagram of a flexible job shop scheduling dual-network collaborative optimization system provided by an embodiment of the present application is exemplified. DETAILED DESCRIPTION
[0044] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.
[0045] To further illustrate the technical solution provided by the embodiment of the present application, this is described in detail below in conjunction with the accompanying drawings and specific implementation methods. Although the embodiment of the present application provides the method operation steps shown in the following embodiments or drawings, more or fewer operation steps may be included in the method based on routine or no creative labor. In the steps where there is no necessary causal relationship logically, the execution order of these steps is not limited to the execution order provided in the embodiment of the present application.
[0046] This application provides a multi-objective optimization method for flow shop group scheduling with a dual evolution mechanism, which includes the following steps:
[0047] S1: Generate an initial solution set using a hybrid method that includes a random partitioning strategy and a heuristic strategy. The heuristic strategy includes a group partitioning strategy GSO, a group sequence initialization strategy GSST, and a workpiece sequence initialization strategy PF.
[0048] Three heuristic methods are used to solve three different sub-problems: (1) Partition the group sequence for in-house processing and outsourcing. (2) Sort the sequence of in-house processing groups. (3) Sort the workpiece sequence within the in-house processing groups, thereby generating an initial population.
[0049] Furthermore, in step S1, the encoding scheme for hybrid initialization includes:
[0050] Partition the group sequence G into in-house processing groups G_in and outsourcing processing groups G_out.
[0051] Partition the workpiece sequence π into in-house processing workpieces π_in and outsourcing workpieces π_out.
[0052] Assign a speed level v to each in-house processing workpiece, and the speed level corresponds to a preset set of machining speed values.
[0053] Furthermore, the group partitioning strategy GSO includes: arranging in ascending order of the total processing time of the workpieces within the group.
[0054] (1) Partition the in-house processing groups and the outsourcing groups. First, use the GSO heuristic algorithm to partition the in-house processing groups and the outsourcing groups. The purpose is to minimize the total cost of outsourcing options and improve production efficiency, thereby optimizing the objective value.
[0055] Based on the optimization objectives of cost and production efficiency, determine the optimal partitioning of the in-house processing groups.
[0056] Furthermore, after partitioning the group sequence into outsourcing and in-house processing, use the GSST (Group Sequence Setup Time) strategy to initialize the in-house group sequence.
[0057] The group sequence initialization strategy GSST based on the setup time between groups includes: constructing a matrix of the total setup time between groups.
[0058] Select the group sequence with the minimum setup time to generate the initial sorting of the in-house processing groups.
[0059] Furthermore, the workpiece sequence initialization strategy PF includes:
[0060] Generate the workpiece processing order by the dynamic insertion method according to the principle of minimizing the total blocking and idle time of the workpieces within the group.
[0061] S2: Divide the initial population into a non-dominated solution set \(P_A\) and a dominated solution set \(P_B\) according to the quality of the solutions.
[0062] S3: Before each iteration, through the pre-evolution evaluation of the temporary population \(P_{tem}\), select the cross-evolution search CES or the improved particle swarm optimization IPSO as the evolution strategy for the non-dominated solution set \(P_A\) and the dominated solution set \(P_B\) respectively.
[0063] The evaluation method for dynamic strategy selection in step S3 is as follows:
[0064] Calculate the fitness value of the solution based on the penalty boundary intersection method PBI, and the formula is:
[0065] f PBI = R1 + θR2.
[0066] Furthermore, the cross-evolution search CES includes: random insertion operation of intra-group workpieces, and group sequence optimization based on block movement search.
[0067] Position-based crossover PBC operation, select the parent solution through roulette wheel and generate the offspring sequence.
[0068] S4: Execute the selected evolution strategy for the non-dominated solution set \(P_A\) and the dominated solution set \(P_B\) respectively, and ensure equilibrium by adjusting the population size.
[0069] In the evolution stage, divide the initial population into two populations according to the dominance relationship between solutions. Among them, the non-dominated solution set is called \(P_A\), and the dominated solution set is called \(P_B\). There are two different evolution strategies for population evolution. One is the cross-evolution search (CES), and the other is the improved strategy based on the particle swarm optimization algorithm (IPSO). Before the start of each iteration, each population will pre-evaluate the evolution strategy suitable for itself, so as to generate more and better solutions during the evolution process. In addition, in order to prevent the scale difference between the non-dominated solution set \(P_A\) and the dominated solution set \(P_B\) of the population from being too large after division, the initial evolution population usually makes the two populations have the same scale by moving some solutions between the two populations.
[0070] (1) Selection of evolution strategy
[0071] Since two populations and two evolutionary strategies are proposed, it is necessary to select which evolutionary strategy for each population to evolve. The purpose is to generate more and better solutions, thus ensuring the convergence and diversity of the final solutions. The evaluation criterion for solution quality uses the penalty-based boundary crossing method (PBI), which is a type of aggregation function commonly used in decomposition-based multi-objective evolutionary algorithms. It mainly uses the vector decomposition method, and the PBI value of each solution can be used to evaluate the quality of the solution. Suppose there are two solutions f1 and f2. If PBI(f1) < PBI(f2), it means that f1 is better than f2. The expression of PBI is the following formula, where θ is the parameter to be adjusted, Z * represents the ideal point, and λ represents the weight vector.
[0072] minf PBI (x|λ,z * ) = R1 + θR2;
[0073]
[0074] R2 = ||F(x) - (z * - R1λ)||;
[0075] Before each iteration, P A and P B will form a temporary population P tem , whose size is NP. P tem randomly selects an evolutionary strategy to evolve, and then sorts the evolved solutions according to the evaluation criterion PBI of the solutions from good to bad, denoted as Select the top NP / 2 individuals from as and compare which of these individuals comes from population P A or P B is more. If the number of P A is more than that of P B , it means that P A is suitable for this strategy, and P A will use this strategy to evolve in the formal evolution. Otherwise, P A will use the other strategy to evolve.
[0076] (2) Cross Evolutionary Search
[0077] The first evolutionary strategy in the evolutionary part is cross evolutionary search (CES). Suppose the population size is P1. Since each solution in the population is divided into an outsourcing sequence in the initialization stage, the CES operation is only performed on the internal group sequence. It includes the random insertion operation of in-group workpieces, the block movement search of the group, and the position-based crossover (PBC) of the group.
[0078] First, perform a random insertion operation on the workpieces within the group to increase the diversity of the solutions. Then, use the block move search on the group. Different from the general "insertion - swap" strategy, since the length of the internal group sequence of each solution is inconsistent, the block size is determined according to the length of the group sequence. This can ensure the diversity of the population and avoid falling into local optima. Finally, use the PBC operation, and the steps are as follows:
[0079] Step 1: Select two internal solution sequences f1 and f2 from the population using the roulette wheel strategy. Since the lengths of the group sequences of the internal schedules are different, it is stipulated to randomly generate a 0 - 1 sequence according to the length of the first group sequence f1.
[0080] Step 2: Take out the groups corresponding to the positions where the elements in the 0 - 1 sequence are 1 from f1 and put them into the new sequence f new .
[0081] Step 3: After removing f from the sequence f1, check whether the remaining elements exist in f2. If they exist, take out these group sequences in the corresponding order in f2 and place them behind f new . new .
[0082] Step 4: If there are still remaining elements in f1 at this time, take out the remaining elements in f1 relative to f new and insert them into the position that can minimize the TEC of f new .
[0083] (3) Improved Particle Swarm Optimization Algorithm
[0084] The second evolutionary strategy is to implement an improved strategy based on the Particle Swarm Optimization Algorithm (IPSO). According to the performance assigned to each particle, the population will be continuously iterated and updated, thus guiding the individuals to develop towards better particles. The Particle Swarm Optimization Algorithm (PSO) is one of the classical swarm intelligence meta - heuristic algorithms and has been widely applied in various evolutionary computing fields. The Particle Swarm Optimization Algorithm with the introduction of the gravity factor can better utilize the individual performance and improve the exploration and exploitation capabilities of the meta - heuristic algorithm, so it is more popular. Combining the characteristics of BGFSP_OO, an improved Particle Swarm Optimization Algorithm (IPSO) is designed, and the main steps are as follows.
[0085] ① Variable Transformation
[0086] Since the particle positions in the Particle Swarm Optimization Algorithm are continuous and cannot be directly used for discrete problems such as BGFSP_OO. Therefore, a transformation method that converts discrete problems into position - continuity problems needs to be designed. Its expression is as follows:
[0087]
[0088] Let \(d\) denote the total number of groups, and \(group(e)\) denote the group index. Each solution sequence is regarded as a particle, and each group within the solution is regarded as a dimension of the particle. Since there are both internally processed groups and outsourced processed groups in each individual, it is assumed that the position information corresponding to the outsourced group sequence in each individual in IPSO is set to 0. \(X_{min}\) and \(X_{max}\) are the lower and upper limits of the position respectively. The discrete information of the group sequence can be converted into continuous position information. On the contrary, after the evolution is completed, the position information needs to be converted into a group sequence, that is, the continuous problem is converted into a discrete problem. The ROV rule can convert the position information into the permutation information of the group sequence. Sort the position information of the particle from small to large, assign the smallest position value as 1, the second smallest position value as 2, and so on, until the position information of the particle is converted into a group sequence. Through these two conversions, the conversion between continuous problems and discrete problems is achieved.
[0089] ② Strategy description
[0090] After completing the conversion of the discrete group sequence into continuous positions, the IPSO strategy is used for the evolution of the particles. The position of each particle \(k\) in the \(t\)-th iteration is defined as follows: The definition expressions for position and velocity are shown as follows.
[0091]
[0092] Among them, \(\gamma\), \(c1\), \(c2\) are set to 1, 2, 2. \(pos\) k is the best position in the history of each particle, and \(b\) is the optimal position of the population. The PBI strategy is used to compare the quality of individuals in the population. \(a\) represents random perturbation, which comes from the perturbation of the particle during the evolution process, aiming to make the particle move towards a better particle.
[0093] S5: Iteratively update the solution set until the termination condition is met, and output the Pareto optimal set of non-dominated solutions.
[0094] On the other hand, the present application also provides a flow shop group scheduling multi-objective optimization system with a double evolution mechanism, which is used to execute the method provided in the first aspect above. It includes:
[0095] An initialization module, which is configured to generate an initial solution set by using a hybrid method including a random partitioning strategy and a heuristic strategy. The heuristic strategy includes a group partitioning strategy GSO, a group sequence initialization strategy GSST, and a workpiece sequence initialization strategy PF.
[0096] A partitioning module, which is configured to partition the initial population into a non-dominated solution set \(P_A\) and a dominated solution set \(P_B\) according to the quality of the solutions.
[0097] A strategy determination module, which is configured to, before each iteration, select cross - evolution search (CES) or improved particle swarm optimization (IPSO) as the evolution strategy for the non - dominated solution set P_A and the dominated solution set P_B respectively through the pre - evolution evaluation of the temporary population P_tem.
[0098] An evolution module, which is configured to execute the selected evolution strategy for the non - dominated solution set P_A and the dominated solution set P_B respectively, and ensure balance by adjusting the population size.
[0099] A Pareto set generation module, which is configured to iteratively update the solution set until the termination condition is met, and output the Pareto - optimal set of non - dominated solutions.
[0100] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer - usable storage media (including but not limited to disk storage, CD - ROM, optical storage, etc.) containing computer - usable program code.
[0101] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general - purpose computer, a special - purpose computer, an embedded processor, or other programmable data - processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data - processing devices generate a device for realizing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.
[0102] These computer program instructions can also be stored in a computer - readable memory that can direct a computer or other programmable data - processing device to work in a specific manner, so that the instructions stored in the computer - readable memory generate a manufactured article including an instruction device, and the instruction device realizes the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.
[0103] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions for implementing the steps specified in one process or a plurality of processes and / or boxes Figure 1 one process or a plurality of processes and / or boxes Figure 1 in one box or a plurality of boxes.
[0104] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application is also intended to include these modifications and variations.
Claims
1. A multi-objective optimization method for flow shop group scheduling with a dual evolution mechanism, characterized in that It includes the following steps: S1: Generate an initial solution set using a hybrid method that includes a random partitioning strategy and a heuristic strategy. The heuristic strategy includes a group partitioning strategy GSO, a group sequence initialization strategy GSST, and a workpiece sequence initialization strategy PF; S2: Divide the initial population into a non-dominated solution set P_A and a dominated solution set P_B according to the quality of the solutions; S3: Before each iteration, through the pre-evolution evaluation of the temporary population P_tem, select the cross-evolution search CES or the improved particle swarm optimization IPSO as the evolution strategy for the non-dominated solution set P_A and the dominated solution set P_B respectively; S4: Execute the selected evolution strategies for the non-dominated solution set P_A and the dominated solution set P_B respectively, and ensure balance by adjusting the population size; S5: Iteratively update the solution set until the termination condition is met, and output the Pareto optimal set of non-dominated solutions.
2. The method according to claim 1, wherein In step S1, the encoding scheme of the hybrid initialization includes: Divide the group sequence G into an internal processing group G_in and an outsourcing processing group G_out; Divide the workpiece sequence π into an internal processing workpiece π_in and an outsourcing workpiece π_out; Assign a speed level v to each internal processing workpiece, and the speed level corresponds to a preset set of machining speed values.
3. The method according to claim 2, characterized in that The group partitioning strategy GSO includes: Arrange in ascending order of the total processing time of the workpieces within the group; Based on the optimization objectives of cost and production efficiency, determine the optimal partitioning of the internal processing group.
4. The method according to claim 1, characterized in that, The group sequence initialization strategy GSST includes: Construct a matrix of the total setup time between groups; Select the group sequence with the minimum setup time to generate the initial sorting of the internal processing group.
5. The method according to claim 4, characterized in that, The workpiece sequence initialization strategy PF includes: Generate the workpiece processing order by the dynamic insertion method according to the principle of minimizing the total blocking and idle time of the workpieces within the group.
6. The method according to claim 1, wherein The evaluation method of the dynamic strategy selection in step S3 is: Calculate the fitness value of the solution based on the penalty boundary cross method PBI, and the formula is: f PBI = R1 + θR2.
7. The method according to claim 1, wherein The cross-evolution search CES includes: Random insertion operation of the workpieces within the group; Group sequence optimization based on block movement search; Position-based crossover PBC operation, select the parent solution through roulette wheel and generate the offspring sequence.
8. A multi-objective optimization system for group scheduling in a flow shop with a dual evolution mechanism, which is used to execute the method described in any one of the above claims 1-7, characterized in that, Includes: Initialization module, which is configured to generate an initial solution set using a hybrid method that includes a random partitioning strategy and a heuristic strategy. The heuristic strategy includes a group partitioning strategy GSO, a group sequence initialization strategy GSST, and a workpiece sequence initialization strategy PF; Partitioning module, which is configured to divide the initial population into a non-dominated solution set P_A and a dominated solution set P_B according to the quality of the solutions; Strategy determination module, which is configured to, before each iteration, through the pre-evolution evaluation of the temporary population P_tem, select the cross-evolution search CES or the improved particle swarm optimization IPSO as the evolution strategy for the non-dominated solution set P_A and the dominated solution set P_B respectively; Evolution module, which is configured to execute the selected evolution strategies for the non-dominated solution set P_A and the dominated solution set P_B respectively, and ensure balance by adjusting the population size; Pareto set generation module, which is configured to iteratively update the solution set until the termination condition is met, and output the Pareto optimal set of non-dominated solutions.