Hybrid flow shop adaptive collaborative optimization method and system

Abstract

The invention provides a hybrid flow shop adaptive collaborative optimization method and system, and belongs to the technical field of workshop production scheduling. The method comprises the following steps: by taking minimum completion time and total energy consumption as objective functions, constructing a hybrid flow shop scheduling model considering grouping setting time and consistent sub-batches, and generating an archive set; the archive set is a reference solution set of the model; a population updating strategy based on fitness and a non-dominated sorting updating strategy guide population updating and archive set updating; the population and the archive set are initialized based on a multi-population adaptive cooperation multi-objective evolutionary algorithm; the weight is dynamically adjusted in different evolution stages by using an adaptive evolution mechanism, and a sub-plot dynamic adjustment energy-saving strategy based on a disjunction graph is introduced, so that finer-grained energy optimization is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of workshop production scheduling, and in particular relates to a method and system for adaptive collaborative optimization of a hybrid flow workshop. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Optimizing productivity and minimizing operating costs in modern manufacturing systems are key to solving production scheduling problems. The Hybrid Flow Shop Scheduling Problem (HFSP) is frequently encountered in high-tech industries such as automotive production, aerospace, and electronics assembly. Unlike the traditional Flow Shop Scheduling Problem (FSP), the HFSP allows for multiple processing stages to be processed in parallel and assigns heterogeneous machines to each stage, providing greater flexibility and adaptability, better reflecting actual production needs.

[0004] In recent years, researchers have proposed methods based on multiple populations and metaheuristic algorithms and have made significant progress. At the same time, the extension problems of HFSP, such as setup time, energy consumption optimization, and batch flow, have also attracted widespread attention. However, existing research still faces the following challenges: First, HFSP is essentially an NP-hard problem. When additional constraints (such as sub-batch consistency, group scheduling, and energy optimization) are added, its complexity will increase significantly; second, most studies are based on ideal assumptions and do not fully consider the multiple coupling constraints in actual production, which limits its practical application; in addition, many studies only focus on a single optimization objective, without optimizing energy consumption and productivity at the same time, which is inconsistent with the trend of green manufacturing and energy conservation and emission reduction. Summary of the Invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a hybrid flow shop adaptive collaborative optimization method and system, which models the hybrid flow shop scheduling problem considering group setting time and consistency sub-batches, adopts a multi-population adaptive collaborative multi-objective evolutionary algorithm to enhance global search capability and local optimization capability, and adopts a dynamic sub-batch adjustment strategy of discrete graph to optimize energy consumption and balance the relationship between productivity and energy efficiency.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: A first aspect of the present invention provides a hybrid flow shop adaptive collaborative optimization method; A hybrid flow shop adaptive collaborative optimization method, comprising: Taking the minimization of completion time and total energy consumption as the objective function, a hybrid flow shop scheduling model considering grouping setup time and consistency sub-batches is constructed to generate an archive set; the archive set is the reference solution set of the model; The fitness-based population update strategy and the non-dominated sorting-based update strategy guide the population update and archive set update; The population and archive set are initialized based on a multi-population adaptive cooperative multi-objective evolutionary algorithm; the adaptive evolutionary mechanism is used to dynamically adjust the weights at different evolutionary stages, and a sub-batch block based on a disjunctive graph is introduced to dynamically adjust the energy-saving strategy to achieve finer-grained energy optimization.

[0007] As a further technical solution, the objective function for minimizing the completion time is: ; Where, is the completion time; is the artifact index, is the total number of workpieces; Index for sub-batch; is the maximum number of sub-batches of workpieces; is the number of parallel machines; is the total number of machines; The objective function of minimizing the total energy consumption is: ; Where, is the total energy consumption; is the stage index; For the stage Machine index on ; For workpiece Sub-batch In the stage Machine Processing time on For speed Unit processing energy consumption under ; For Group To Group In the machine Setting time on Prepare energy consumption for the unit of machine; is the total standby time; It is the unit standby energy consumption of the machine.

[0008] As a further technical solution, the fitness-based population update strategy includes: Calculate the completion time. The completion time is the last time point when all tasks are completed in the entire production scheduling process, as shown in the following formula: ; Where, For workpiece Sub-batch In stage Machine Completion time on is a decision variable, representing the workpiece Sub-batch In stage Machine Upper processing; Calculate total energy consumption including processing, machine setup, and standby energy; First, calculate the waiting time of the entire processing process , processing time and machine setup time , ; Where, is the total number of workpieces; is the maximum number of sub-batches of workpieces; is the number of parallel machines; is the total number of machines; Index for artifacts; Index for sub-batch; is the stage index; For the stage Machine index on ; For workpiece Sub-batch In stage Machine Processing time on For workpiece Sub-batch In stage Machine Upper processing; Calculate total energy consumption ,in represents the processing energy consumption, Indicates the energy consumption of the machine settings. Indicates the energy consumption of the machine in standby mode; ; ; ; ; Where, For speed Unit processing energy consumption under ; For workpiece Sub-batch In stage Machine Speed grade processing; is the group index, ; is the total number of groups; is the number of parallel machines; is the total number of machines; For Group To Group In the machine Setting time on For the stage Up, Group In group Process immediately afterwards; Prepare energy consumption for the unit of machine; is the total standby time; is the unit standby energy consumption of the machine; The performance index of each solution is calculated by weighted average method ; Weight factor Used to balance between completion time and energy consumption; ; The strategy is updated based on the population update and the archive set update; if the performance of the current solution is better than the solution in the population, the solution in the population is updated.

[0009] As a further technical solution, the update strategy based on non-dominated sorting includes a diversity preservation strategy based on non-dominated sorting and crowding distance and a diversity preservation strategy based on non-dominated sorting and Euclidean distance.

[0010] As a further technical solution, the diversity preservation strategy based on non-dominated sorting and crowding distance combines non-dominated sorting and crowding distance to maintain population diversity. The specific steps are as follows: (1) Assume that there are multiple solutions, each with a fitness value; set a predefined number N of solutions to be selected, and select a specific number of solutions from the set; (2) Perform non-dominated sorting on the solution set and divide the solutions into different frontier layers; (3) For each frontier layer, calculate the crowding distance of the solution in the target space; (4) Select solutions from each frontier layer, giving priority to solutions with larger crowding distances; if the total number of selected solutions plus the number of currently selected solutions exceeds the predefined N, filter solutions from the current layer according to crowding distances and retain the solution with the largest crowding distance.

[0011] As a further technical solution, the diversity preservation strategy based on non-dominated sorting and Euclidean distance combines non-dominated sorting and Euclidean distance to select a solution from the solution set; the specific steps are as follows: (1) Assume that there are multiple solutions, each with a corresponding fitness value; set a predefined number N of solutions to be selected, and select a specific number of solutions from them; (2) Perform non-dominated sorting on the solution set and divide the solutions into different frontier layers; (3) For each frontier layer, if the number of solutions exceeds the predefined maximum selection number n, calculate the Euclidean distance between the solutions and select the solution with the farthest distance; (4) If the number of selected solutions exceeds the predefined number N, the Euclidean distance is further used to retain the solution with the highest diversity.

[0012] As a further technical solution, the process of initializing the population based on a multi-population adaptive cooperative multi-objective evolutionary algorithm includes: Copy sequences directly from an archive set ; Randomly select sequences from the archive set , and perturbs them by group random exchange, group random reversal and group random recombination operators; The remaining sequences were generated based on random replication or probabilistic selection.

[0013] As a further technical solution, the process of dynamically adjusting weights at different evolutionary stages using the adaptive evolution mechanism is as follows: An activation function based on solution set characteristics is introduced. The activation function calculates the coefficient of variation based on the completion time and energy consumption of the non-dominated solutions in the solution set, and combines the median congestion distance of the non-dominated solutions to evaluate the status of the current solution set. Adaptive evolution is performed by iteratively improving the optimization group population, artifact population, and artifact segmentation population, and weights are adaptively adjusted according to the optimization progress; specifically: The average fitness value of the new population, the average fitness value of the initialized population and the average fitness value before the last weight adjustment are extracted respectively, and the weights are adjusted by calculating the relative improvement of completion time and energy consumption.

[0014] A second aspect of the present invention provides a hybrid flow shop adaptive collaborative optimization system.

[0015] A hybrid flow shop adaptive collaborative optimization system, comprising: The scheduling model building module is configured to: construct a hybrid flow shop scheduling model that considers grouping setup time and consistency sub-batches, taking minimization of completion time and total energy consumption as an objective function, and generate an archive set; the archive set is a reference solution set of the model; The population update module is configured to: guide the population update and archive set update based on the fitness-based population update strategy and the non-dominated sorting-based update strategy; The optimization module is configured to: initialize the population and archive set based on a multi-population adaptive collaborative multi-objective evolutionary algorithm; use the adaptive evolution mechanism to dynamically adjust the weights at different evolutionary stages, and introduce a sub-plot dynamic adjustment energy-saving strategy based on a disjunctive graph to achieve finer-grained energy optimization.

[0016] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of the hybrid flow shop adaptive collaborative optimization method as described in the first aspect of the present invention.

[0017] The fourth aspect of the present invention provides an electronic device, comprising a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of the method for adaptive collaborative optimization of a hybrid flow shop as described in the first aspect of the present invention are implemented.

[0018] One or more of the above technical solutions have the following beneficial effects: This paper integrates heuristic rules and randomized strategies to construct a high-quality, diverse initial solution set. A multi-level co-evolutionary mechanism is then designed, with specific operators developed for each level to exploit the potential for collaboration between populations, thereby enhancing both global search and local exploitation capabilities. To further optimize search efficiency, an adaptive evolutionary mechanism is proposed that dynamically adjusts weights at different evolutionary stages to balance exploration and exploitation, effectively avoiding local optima. Furthermore, a disjunctive graph-based sub-plot dynamic adjustment energy-saving strategy is introduced to achieve finer-grained energy optimization.

[0019] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0021] Figure 1 This is a flow chart of the method of the first embodiment.

[0022] Figure 2 This is a system structure diagram of the second embodiment. DETAILED DESCRIPTION

[0023] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0024] It should be noted that the terms used herein are for describing particular embodiments only and are not intended to limit the exemplary embodiments according to the present invention.

[0025] In the absence of conflict, the embodiments of the present invention and the features thereof may be combined with each other.

[0026] Example 1 This embodiment discloses a hybrid flow shop adaptive collaborative optimization method. In this embodiment, for a hybrid flow shop problem including group scheduling and batch processing, in this problem, a set of δ groups , where each group Contains multiple artifacts Each group The artifacts in the can be further split into multiple sub-batches, each of which must go through The processing stages are represented by At each stage, there are parallel machines are available, represented by , and each machine has different processing speed and capacity. Based on the characteristics of GHFSP_CS, the goal is to find the optimal solution for the following aspects: (1) the order of grouping; (2) the order of workpieces within each group; (3) the allocation of sub-batches; (4) the size of sub-batches; (5) the machine allocation for each sub-batch in each stage; (6) the speed setting of all machines processing sub-batches. The goal of the problem is to find a solution that makes the completion time ( ) and total energy consumption (TEC) are minimized.

[0027] like Figure 1 As shown, a hybrid flow shop adaptive collaborative optimization method includes: Step S1, taking the completion time and the minimum total energy consumption as the objective function, a hybrid flow shop scheduling model considering the grouping setup time and consistency sub-batches is constructed; Among them, the objective function of minimizing the completion time is: ; Where, is the completion time; j is the workpiece index, n is the total number of workpieces; e is the sub-batch index. L is the maximum number of sub-batches of workpieces; m is the number of parallel machines; k is the total number of machines; The objective function of minimizing the total energy consumption is: ; Where, is the total energy consumption; is the stage index; For the stage Machine index on ; For workpiece Sub-batch In stage Machine Processing time on For speed Unit processing energy consumption under ; For Group To Group In the machine Setting time on Prepare energy consumption for the unit of machine; is the total standby time; is the unit standby energy consumption of the machine. The constraints of the above objective function are set as: (1) , , which means that each artifact has one and only one direct predecessor artifact and one direct successor artifact; (2) , It means that each group has a unique predecessor group and a unique successor group; (3) , which means that each sub-batch of workpieces can only be processed by one machine at each stage; (4) This means that each machine can only process one sub-batch at each stage at the same time; (5) This means that each sub-batch can only select one speed level for processing at each stage; (6) Used to define the processing time of sub-batches; (7) Used to ensure that the total size of all subbatches is equal to the total number of artifacts; (8) Used to assign each artifact to a specific group; (9) Represents the relationship between an artifact and its grouping; (10) Indicates that the start time of a sub-batch is no earlier than its completion time on the machine selected in the previous stage; (11) It represents the association between two consecutive workpieces in the same group, at the same stage, and on the same machine; (12) It represents the association between two consecutive workpieces in different groups at the same stage and on the same machine; (13) Indicates the completion time of the first sub-batch to be processed; (14) Indicates the standby time of the entire process.

[0028] Step S2, guiding population update based on fitness population update strategy and non-dominated sorting update strategy; Among them, for the population update strategy based on fitness, since the research goal is to optimize the completion time and energy consumption, the fitness evaluation is based on the and The fitness evaluation process can be divided into the following key steps: Step S21, calculate completion time Completion time is the final time point when all tasks are completed in the entire production scheduling process.

[0029] ; Step S22, calculate energy consumption The total energy consumption including processing, machine setup and standby energy consumption was calculated. The specific calculation process is as follows: Calculate the standby time of the entire processing process , processing time and machine setup time .

[0030] ; ; Calculate total energy consumption ,in represents the processing energy consumption, Indicates the energy consumption of the machine settings. Indicates the energy consumption of the machine in standby mode.

[0031] ; ; ; .

[0032] Step S23, calculating performance indicators In this step, the performance index of each solution is calculated by the weighted average method . Weighting factor Used to balance between completion time and energy consumption.

[0033] ; Step S24, update strategy Update strategies include population updates and archive set updates.

[0034] If the current solution outperforms the solutions in the population and the collaborators have not changed, the solution in the population is updated. If a part of the solution has changed, the solution is updated regardless of its performance.

[0035] The archive update rule depends on the performance improvement, and a new solution is added to the archive only if it outperforms the existing solutions in terms of fitness and shows a greater improvement among other new solutions, ensuring that the optimal solutions are retained in the archive.

[0036] For the update strategy based on non-dominated sorting, in order to prevent premature convergence to the local optimum and maintain population diversity, this embodiment proposes two strategies based on non-dominated sorting to maintain diversity.

[0037] For the diversity preservation strategy based on non-dominated sorting and crowding distance, this strategy combines non-dominated sorting and crowding distance to maintain population diversity. The specific steps are as follows: (1) Assume that there are multiple solutions, each with a fitness value. Set a predefined number N of solutions to be selected, and select a specific number of solutions from the set.

[0038] (2) Perform non-dominated sorting on the solution set and divide the solutions into different frontier layers.

[0039] (3) For each frontier layer, calculate the crowding distance of the solution in the target space.

[0040] (4) Select solutions from each frontier layer, giving priority to solutions with larger crowding distances. If the total number of selected solutions plus the number of currently selected solutions exceeds the predefined N, filter solutions from the current layer according to crowding distances and retain the solution with the largest crowding distance.

[0041] For the diversity preservation strategy based on non-dominated sorting and Euclidean distance, this strategy combines non-dominated sorting and Euclidean distance to select a solution from the solution set. The specific steps are as follows: (1) Assume that there are multiple solutions, each with a corresponding fitness value. Set a predefined number N of solutions to be selected, and select a specific number of solutions from them.

[0042] (2) Perform non-dominated sorting on the solution set and divide the solutions into different frontier layers.

[0043] (3) For each frontier layer, if the number of solutions exceeds the predefined maximum selection number n, the Euclidean distance between the solutions is calculated and the solution with the farthest distance is selected.

[0044] (4) If the number of selected solutions exceeds the predefined number N, the Euclidean distance is further used to retain the solution with the highest diversity. In step S3, the initial solution set of the scheduling model is obtained based on a multi-population adaptive cooperative multi-objective evolutionary algorithm. The archive set is initialized, the weights are dynamically adjusted at different evolutionary stages using an adaptive evolutionary mechanism, and a sub-plot dynamic adjustment energy-saving strategy based on a disjunctive graph is introduced to achieve finer-grained energy optimization.

[0045] Among them, the multi-objective evolutionary algorithm based on multi-population adaptive cooperation consists of four parts: initialization, multi-population collaborative evolution, adaptive evolution and energy-saving strategy. These four structures enable the algorithm to fully explore the solution space through interaction and information sharing. The archive set contains reference solutions, which serve as external memory to save the optimal solution found during the search process and guide the evolution of the population throughout the search process. In addition, you can see the grouping sequence , artifact sequence , and the artifact splitting matrix From the archive set, these can constitute a complete solution in the group population.

[0046] The solution to the GHFSP_CS problem can be represented by a five-tuple Represents, where the first element is a grouped sequence The second element is the sequence of artifacts within each group ,in Indicates grouping The processing sequence of the internal workpiece, . is the machine assignment matrix of subbatches on each stage. Used to determine the allocation of subbatches and the size of each subbatch of artifacts. is the matrix of processing speed ratings for each sub-batch on the selected machines. The encoding vector defines the grouping order and the order of workpieces within each group in the first stage. In subsequent stages, workpieces are scheduled based on their completion time from the previous stage. Machine assignment follows the earliest completion time (ECT) rule.

[0047] Two heuristics are used to initialize the archive set to match the complexity of the problem. The first method (LHS-MPS-GS) generates solutions, where LHS is used for the artifact splitting matrix, MPS is used for the artifact sequence, and GS is used for the grouping sequence. The high-quality solutions are then identified by non-dominated sorting. The second method (SGS-RAN) generates Solutions are generated by combining SGS (Sequential Generation Strategy) and random strategy. The remaining solutions are generated randomly.

[0048] The group population is initialized in three ways. First, the sequences are copied directly from the archive set. . Second, randomly select sequences from the archive set , and is perturbed by group random swap (GR-S), group random reversal (GR-R), and group random recombination (GR-O) operators. Finally, the remaining sequence is generated based on random replication or probabilistic selection. In this process, a complete solution also needs to include the collaborative artifact sequence and split matrix These are initialized via direct copying, random selection, or perturbation strategies.

[0049] Similarly, the artifact population and artifact split population contain solutions, each of which can be expressed as For the partial solutions in the workpiece population and the workpiece split population ,The group sequences are both generated by random perturbations, with the difference that the ,artifact segmentation matrix in the artifact segmentation group is randomly ,drawn from the archive set and obtained by applying the random row ,swap (RRS) operation, while the artifact segmentation matrix in the ,artifact group is simply randomly drawn from the archive set.,Each component of the solution is randomly ,drawn from the archive set.

[0050] During the population evolution process, the archive set is divided into the Pareto front set and the non-Pareto frontier set For each group sequence , their collaborative artifact sequence and split matrix The job partitioning group is selected from or based on probability. To improve diversity and stability, operations such as greedy block swap mutation (G-GBSM), group greedy insertion (G-GI), and group reversal (GR) are applied. If a new solution improves fitness, it replaces the old one and updates the archive. The evolution of the job partitioning group is similar, except that mutation greedy block swap (S-GBS), boundary insertion (S-BI), and subslot mutation (S-SM) operations are performed on the job partitioning matrix.

[0051] During the evolution of the artifact population, random mapping , A cooperative group sequence will be generated and the artifact partition matrix ,Two-point sequence crossover (J-TPC) is applied to the artifact sequence to construct a new solution, which is then updated using the same mechanism as for the group.

[0052] In the early evolution process, fixed weights were used To update the solution, we need to ensure the stability of the optimization process. However, in this part, we design an adaptive evolutionary strategy to dynamically adjust the population. This strategy includes calculating the fitness of the generated solution in the co-evolution, determining the activation conditions of the adaptive evolution, and executing a dynamic weight adjustment mechanism when the conditions are met.

[0053] In order to improve the adaptability of the algorithm, an activation function based on the solution set characteristics is introduced to determine when to trigger the adaptive evolution mechanism. Specifically, the activation function calculates the coefficient of variation based on the completion time and energy consumption of the non-dominated solutions in the solution set. (Coefficient of Variation, CV), and combined with the median of the crowding distance of the non-dominated solution to evaluate the status of the current solution set.

[0054] The activation function sets the following conditions: when the coefficient of variation of both the completion time and energy consumption falls below a preset threshold, or when the median crowding distance of the non-dominated solutions falls below a set value, the adaptive evolution mechanism is activated. This design ensures that if the solution set is too convergent or lacks diversity, the algorithm can adjust the population appropriately, thereby escaping local optima and maintaining its ability to explore the solution set.

[0055] ; ; in, represent the coefficient of variation of completion time and energy consumption respectively, is the standard deviation of completion time and energy consumption, is the mean of completion time and energy consumption.

[0056] The adaptive evolution process aims to iteratively improve the optimization group population, artifact population, and artifact split population. Its goal is to adaptively adjust weights based on the optimization progress to improve diversity, stability, and solution quality. To this end, specific operators are designed for each population: G-GS, J-BS, and S-RGI. The weight adjustment rules are as follows: (1) Extract the average value of the new population fitness value Assume that the new population before the last weight adjustment includes solutions, whose fitness values are , the average fitness value of the new population can be calculated as follows: ; in , They represent the average fitness values of completion time and energy consumption in the new population respectively.

[0057] (2) Extract the average of the fitness values of the initialized population and before the last weight adjustment If this is the first time to adjust the weights, then only the fitness mean of the initialized population needs to be calculated. If this is not the first time, then the fitness mean of the new population solution before the last adjustment and the initialized population solution should be calculated. Assume that the total number of these solutions is , and the fitness values are , we can get the average of the fitness values of the initialized population and before the last weight adjustment, as shown below: ; in , They represent the average fitness values of the completion time and energy consumption of the initialized population and the new population before the last adjustment, respectively.

[0058] (3) Calculate the relative improvement Calculate the relative improvement of completion time and energy consumption respectively, using 、 express: ; ; (4) Adjust weights

[0059] if ,but Decrease by 0.1, otherwise Increase by 0.1.

[0060] Energy-saving strategy based on disjunctive graph In multi-stage shop floor scheduling, processing speed significantly impacts energy consumption. This paper proposes a dynamic sub-batch adjustment energy-saving strategy based on disjunctive graphs. This strategy optimizes sub-station processing speed by analyzing the dependencies and resource competition between conjunction and disjunctive arcs. Each node represents a task, containing processing time, speed, and dependency information. Nodes of the same color share the same machine at a specific stage.

[0061] The strategy optimization targets subbatches on non-critical paths and exploits time redundancy by locating critical nodes and optimizing their processing speed. The optimization is subject to the following strict constraints: (1) Subbatch successor constraint: the completion time must not be later than the start time of the subsequent node; (2) Machine successor constraint: it must not affect the execution of subsequent nodes on the same machine; and (3) Global scheduling performance constraint: the completion time must not increase after the adjustment. The adjustment prioritizes energy-efficient processing speed and uses an incremental update mechanism to iteratively improve the solution. After implementing the energy-saving strategy, the processing time of some subbatches increased, but it did not affect the processing start time of their subsequent subbatches and machines. In this way, the completion time of the overall process was not affected and energy consumption was reduced, verifying the effectiveness of this method in optimizing energy consumption.

[0062] Example 2 This embodiment discloses a hybrid flow shop adaptive collaborative optimization system; like Figure 2 As shown, a hybrid flow shop adaptive collaborative optimization system includes: The scheduling model building module is configured to: construct a hybrid flow shop scheduling model that considers grouping setup time and consistency sub-batches, taking minimization of completion time and total energy consumption as an objective function, and generate an archive set; the archive set is a reference solution set of the model; The population update module is configured to: guide the population update and archive set update based on the fitness-based population update strategy and the non-dominated sorting-based update strategy; The optimization module is configured to: initialize the population and archive set based on a multi-population adaptive collaborative multi-objective evolutionary algorithm; use the adaptive evolution mechanism to dynamically adjust the weights at different evolutionary stages, and introduce a sub-plot dynamic adjustment energy-saving strategy based on a disjunctive graph to achieve finer-grained energy optimization.

[0063] Example 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0064] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the hybrid flow shop adaptive collaborative optimization method as described in Example 1.

[0065] Example 4 The purpose of this embodiment is to provide an electronic device.

[0066] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of the method for adaptive collaborative optimization of a hybrid flow shop as described in Example 1 are implemented.

[0067] The steps involved in the apparatuses of Examples 2, 3, and 4 above correspond to those of Method Example 1. For detailed implementations, please refer to the relevant description of Example 1. The term "computer-readable storage medium" should be understood to mean a single medium or multiple media containing one or more instruction sets; it should also be understood to include any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and causing the processor to perform any method of the present invention.

[0068] Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using a general-purpose computer device. Alternatively, they can be implemented using program code executable by a computing device, which can then be stored in a storage device and executed by the computing device. Alternatively, they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0069] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A hybrid flow shop adaptive collaborative optimization method, characterized by: include: Taking the minimization of completion time and total energy consumption as the objective function, a hybrid flow shop scheduling model considering grouping setup time and consistency sub-batches is constructed to generate an archive set; the archive set is the reference solution set of the model; The fitness-based population update strategy and the non-dominated sorting-based update strategy guide the population update and archive set update; The population and archive set are initialized based on a multi-population adaptive cooperative multi-objective evolutionary algorithm; the adaptive evolutionary mechanism is used to dynamically adjust the weights at different evolutionary stages, and a sub-batch block based on a disjunctive graph is introduced to dynamically adjust the energy-saving strategy to achieve finer-grained energy optimization.

2. The hybrid flow shop adaptive collaborative optimization method according to claim 1, characterized in that: The objective function of minimizing the completion time is: ； Where, is the completion time; is the artifact index, is the total number of workpieces; Index for sub-batch; is the maximum number of sub-batches of workpieces; is the number of parallel machines; is the total number of machines; The objective function of minimizing the total energy consumption is: ； Where, is the total energy consumption; is the stage index; For the stage Machine index on ; For workpiece Sub-batch In stage Machine Processing time on For speed Unit processing energy consumption under For Group To Group In the machine Setting time on Prepare energy consumption for the unit of machine; is the total standby time; It is the unit standby energy consumption of the machine.

3. The hybrid flow shop adaptive collaborative optimization method according to claim 1, characterized in that: The fitness-based population update strategy includes: Calculate the completion time. The completion time is the last time point when all tasks are completed in the entire production scheduling process, as shown in the following formula: ； Where, For workpiece Sub-batch In stage Machine Completion time on is a decision variable, representing the workpiece Sub-batch In stage Machine Upper processing; Calculate total energy consumption including processing, machine setup, and standby energy; First, calculate the waiting time of the entire processing process , processing time and machine setup time , ； Where, is the total number of workpieces; is the maximum number of sub-batches of workpieces; is the number of parallel machines; is the total number of machines; Index for artifacts; Index for sub-batch; is the stage index; For the stage Machine index on ; For workpiece Sub-batch In stage Machine Processing time on For workpiece Sub-batch In stage Machine Upper processing; Calculate total energy consumption ,in represents the processing energy consumption, Indicates the energy consumption of the machine settings. Indicates the energy consumption of the machine in standby mode; ； ； ； ； Where, For speed Unit processing energy consumption under For workpiece Sub-batch In stage Machine Speed grade processing; is the group index, ; is the total number of groups; is the number of parallel machines; is the total number of machines; For Group To Group In the machine Setting time on For the stage Up, Group In group Process immediately afterwards; Prepare energy consumption for the unit of machine; is the total standby time; is the unit standby energy consumption of the machine; The performance index of each solution is calculated by weighted average method ; Weight factor Used to balance between completion time and energy consumption; ； Update the strategy based on population update and archive set update; if the performance of the current solution is better than the solution in the population, update the solution in the population; As a further technical solution, the update strategy based on non-dominated sorting includes a diversity preservation strategy based on non-dominated sorting and crowding distance and a diversity preservation strategy based on non-dominated sorting and Euclidean distance.

4. The hybrid flow shop adaptive collaborative optimization method according to claim 3, characterized in that: The diversity preservation strategy based on non-dominated sorting and crowding distance combines non-dominated sorting and crowding distance to maintain population diversity. The specific steps are as follows: Assume that there are multiple solutions, each with a fitness value; set a predefined number N of solutions to be selected, and select a specific number of solutions from the set; Perform non-dominated sorting on the solution set and divide the solutions into different frontier layers; For each frontier layer, calculate the crowding distance of the solution in the target space; Solutions are selected from each frontier layer, with priority given to solutions with larger crowding distances. If the total number of selected solutions plus the number of currently selected solutions exceeds the predefined N, solutions are filtered from the current layer according to crowding distances, and the solution with the largest crowding distance is retained.

5. The hybrid flow shop adaptive collaborative optimization method according to claim 3, characterized in that: The diversity preservation strategy based on non-dominated sorting and Euclidean distance combines non-dominated sorting and Euclidean distance to select a solution from the solution set; the specific steps are as follows: Assume that there are multiple solutions, each with a corresponding fitness value; set a predefined number N of solutions to be selected, and select a specific number of solutions from them; Perform non-dominated sorting on the solution set and divide the solutions into different frontier layers; For each frontier layer, if the number of solutions exceeds the predefined maximum selection number n, the Euclidean distance between the solutions is calculated and the solution with the farthest distance is selected; If the number of selected solutions exceeds the predefined number N, the Euclidean distance is further used to retain the solution with the highest diversity.

6. The hybrid flow shop adaptive collaborative optimization method according to claim 1, characterized in that: The process of population initialization based on multi-population adaptive cooperative multi-objective evolutionary algorithm includes: Copy sequences directly from an archive set ; Randomly select sequences from the archive set , and perturbs them by group random exchange, group random reversal and group random recombination operators; The remaining sequences were generated based on random replication or probabilistic selection.

7. The hybrid flow shop adaptive collaborative optimization method according to claim 1, characterized in that: The process of dynamically adjusting weights at different evolutionary stages using the adaptive evolution mechanism is as follows: An activation function based on solution set characteristics is introduced. The activation function calculates the coefficient of variation based on the completion time and energy consumption of the non-dominated solutions in the solution set, and combines the median congestion distance of the non-dominated solutions to evaluate the status of the current solution set. Adaptive evolution is performed by iteratively improving the optimization group population, artifact population, and artifact segmentation population, and weights are adaptively adjusted according to the optimization progress; specifically: The average fitness value of the new population, the average fitness value of the initialized population and the average fitness value before the last weight adjustment are extracted respectively, and the weights are adjusted by calculating the relative improvement of completion time and energy consumption.

8. A hybrid flow shop adaptive collaborative optimization system, characterized by: include: The scheduling model building module is configured to: construct a hybrid flow shop scheduling model that considers grouping setup time and consistency sub-batches, taking minimization of completion time and total energy consumption as an objective function, and generate an archive set; the archive set is a reference solution set of the model; The population update module is configured to: guide the population update and archive set update based on the fitness-based population update strategy and the non-dominated sorting-based update strategy; The optimization module is configured to: initialize the population and archive set based on a multi-population adaptive collaborative multi-objective evolutionary algorithm; use the adaptive evolution mechanism to dynamically adjust the weights at different evolutionary stages, and introduce a sub-plot dynamic adjustment energy-saving strategy based on a disjunctive graph to achieve finer-grained energy optimization.

9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps of the hybrid flow shop adaptive collaborative optimization method according to any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps in the hybrid flow shop adaptive collaborative optimization method according to any one of claims 1 to 7 are implemented.

Images