Method for mixed flow shop scheduling based on INSGA-II and application

Abstract

The application discloses a mixed flow shop scheduling method based on INSGA-II and application. The mixed flow shop scheduling method comprises the following steps: obtaining processing data, establishing a multi-objective optimization scheduling model; generating an initial population by adopting double-layer encoding / decoding rules; performing cross operation and mutation operation; performing selection operation to reserve a next generation population; performing neighborhood search to expand the next generation population; and repeatedly performing iteration update until a termination condition is met. The application avoids the problem of weight assignment in solving a multi-objective optimization problem by adopting a linear weighted summation method in the prior art, and multiple non-inferior solutions can be obtained each time for a decision maker to select. The application performs neighborhood search on population individuals after each iteration, and different search rules and neighborhood structures are designed for workpiece codes and machine codes, so that the local search capability of the algorithm is improved, and the solution quality is improved.

Description

Technical Field

[0001] This invention relates to the field of scheduling optimization and intelligent decision-making technology, and in particular to a hybrid flow shop scheduling method and application based on INSGA-II. Background Technology

[0002] With the rapid development of China's manufacturing industry in recent years, market competition has become increasingly fierce, and market demand has become more diversified. Production methods are gradually shifting from traditional large-scale mass production to multi-variety, small-batch production. Hybrid assembly line scheduling systems offer greater adaptability and flexibility, enabling them to meet the needs of enterprises in response to changes in the external environment.

[0003] The Hybrid Flow-shop Scheduling Problem (HFSP), as an extension of the traditional flow-shop scheduling problem, has been proven to be an NP-hard problem. Compared with the traditional flow-shop scheduling problem, the HFSP problem has higher complexity, and the traditional manual scheduling method can no longer meet the current production needs of enterprises.

[0004] Currently, commonly used methods for solving the High-Speed ​​Workshop Problem (HFSP) are mainly divided into exact algorithms and approximate algorithms. Exact algorithms mainly include branch and bound, dynamic programming, Lagrange relaxation, and integer programming. These algorithms can find exact solutions to problems that are not complex and are small in scale, but their search efficiency is low for large-scale problems with large solution spaces, making it difficult to find the optimal solution. Approximate algorithms are further divided into heuristic algorithms and metaheuristic algorithms. Heuristic algorithms are simple in principle and easy to implement, but they suffer from poor global optimization ability and difficulty in guaranteeing solution quality. Metaheuristic algorithms introduce random factors during the search process and repeatedly call the guiding algorithm to generate better solutions, ensuring that they can find the optimal solution to the problem quickly and stably. Among the many metaheuristic algorithms, genetic algorithms have a significant advantage in solving the workshop scheduling problem due to their strong global search ability and parallelism.

[0005] Some existing technologies, such as Chinese invention patent CN110619437A, provide a low-energy flexible job shop scheduling method based on an improved genetic algorithm. The implementation process of this method is as follows: Figure 1 As shown. Specifically includes:

[0006] Step 1: Construct a workshop scheduling model with energy consumption and completion time as optimization objectives;

[0007] Step 2: Transform the multi-objective optimization problem into a single-objective optimization problem using the linear weighted summation method;

[0008] Step 3: Design a multi-layered encoding method and randomly initialize the population;

[0009] Step 4: Take the weighted summation of the objective function as the fitness function and calculate the fitness value;

[0010] Step 5: Based on the fitness value, select individuals using a roulette wheel method, in preparation for crossover and mutation;

[0011] Step 6: Perform the crossover operation using the integer crossover method;

[0012] Step 7: Mutation operation;

[0013] Step 8: Determine if the maximum number of iterations has been reached. If yes, output the current optimal solution; otherwise, go to step 4.

[0014] However, the inventors of this invention have discovered through long-term practice that in the aforementioned prior art, the improved genetic algorithm is used to solve the low-energy flexible job shop scheduling problem. Specifically, the multi-objective optimization problem is first transformed into a single-objective optimization problem using a linear weighted summation method, and then solved using the improved genetic algorithm. When using the linear weighted summation method, it is necessary to determine the weight of each optimization objective. Decision-makers typically assign these weights manually based on their own needs and historical experience, resulting in problems such as unreasonable weight assignments and poor objectivity. Furthermore, this scheduling method can only obtain one solution at a time, and its performance in practical applications is not ideal. Summary of the Invention

[0015] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a hybrid flow workshop scheduling method and application based on INSGA-II.

[0016] To achieve the aforementioned objectives, the technical solution adopted by this invention includes:

[0017] In a first aspect, the present invention provides a hybrid flow shop scheduling method based on INSGA-II, comprising:

[0018] Acquire processing data and establish a multi-objective optimization scheduling model, wherein the optimization objectives in the multi-objective optimization scheduling model include completion time, energy consumption, and total load;

[0019] The initial population is generated based on a two-layer encoding / decoding rule. The gene encoding generated by the two-layer encoding / decoding rule includes a workpiece code and a machine code. The workpiece code represents the processing order of each workpiece in the first process, and the machine code represents the selection of the processing machine in each process.

[0020] Crossover and mutation operations are performed on the chromosomes in the initial population, and the crossover and / or mutation operations employ different operation methods for the work code and machine code of the chromosomes;

[0021] A selection operation is performed on the chromosomes, retaining the portion of chromosomes that are close to the optimization target as the next generation population;

[0022] A neighborhood search is performed on the next generation population to expand the next generation population;

[0023] Using the next generation population as the new initial population, the iterative update is repeated until the termination condition is met.

[0024] Secondly, the present invention also provides a hybrid assembly line scheduling system based on INSGA-II, comprising:

[0025] The model building module is used to acquire processing data and build a multi-objective optimization scheduling model. The optimization objectives in the multi-objective optimization scheduling model include completion time, energy consumption, and total load.

[0026] The population generation module is used to generate an initial population based on a two-layer encoding / decoding rule. The gene encoding generated by the two-layer encoding / decoding rule includes a workpiece code and a machine code. The workpiece code represents the processing order of each workpiece in the first process, and the machine code represents the selection of the processing machine in each process.

[0027] The crossover and mutation module is used to perform crossover and mutation operations on chromosomes in the initial population. The crossover and / or mutation operations employ different operation methods for the work code and machine code of the chromosomes.

[0028] The selection and retention module is used to select chromosomes and retain a portion of chromosomes that are close to the optimization target as the next generation population.

[0029] The neighborhood search module is used to perform a neighborhood search on the next generation population in order to expand the next generation population;

[0030] The iterative update module is used to repeatedly perform iterative updates using the next generation population as the new initial population until the termination condition is met.

[0031] Thirdly, the present invention also provides a readable storage medium storing a computer program, which, when run, executes the steps of the above-described hybrid assembly line scheduling method.

[0032] Based on the above technical solution, compared with the prior art, the beneficial effects of the present invention include at least the following:

[0033] 1. Compared with the prior art, the multi-objective scheduling method based on INSGA-II proposed in this invention avoids the problem of weight assignment in solving multi-objective optimization problems using the linear weighted summation method, and can obtain multiple non-dominated solutions for decision-makers to choose from each solution.

[0034] 2. In the INSGA-II algorithm proposed in this invention, a neighborhood search is performed on each individual in the population after each iteration of merging. Different search rules and neighborhood structures are designed for work code and machine code respectively, thereby improving the local search capability of the algorithm and improving the quality of the solution.

[0035] The above description is merely an overview of the technical solution of the present invention. In order to enable those skilled in the art to better understand the technical means of this application and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described below in conjunction with detailed drawings. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the process flow diagram of the hybrid assembly line scheduling method provided in the background technology of this invention;

[0037] Figure 2 This is a flowchart illustrating a hybrid assembly line scheduling method provided in a typical embodiment of the present invention.

[0038] Figure 3 This is a schematic diagram of the chromosome coding process of the hybrid flow workshop scheduling method provided in a typical embodiment of the present invention;

[0039] Figure 4 This is a schematic diagram of the non-dominated sorting calculation process of the hybrid flow workshop scheduling method provided in a typical embodiment of the present invention;

[0040] Figure 5 This is a schematic diagram of the neighborhood structure of the hybrid flow workshop scheduling method provided in a typical embodiment of the present invention. Detailed Implementation

[0041] In view of the shortcomings of the prior art, the inventors of this invention, through long-term research and extensive practice, have proposed the technical solution of this invention. The following will further explain and illustrate this technical solution, its implementation process, and its principles.

[0042] To address the High-Stakes Workflow Problem (HFSP), this invention proposes a multi-objective optimization scheduling method based on the Improved Non-dominated Sorting Genetic Algorithm-II (INSGA-II). A multi-objective scheduling model for a hybrid flow shop is established, with the optimization objectives of minimizing completion time, energy consumption, and total load. A multi-objective scheduling method based on INSGA-II is also proposed. Compared to the problems of unreasonable weight assignment and poor objectivity encountered by traditional genetic algorithms in solving multi-objective optimization problems, this method avoids the weight assignment problem and can obtain multiple non-dominated solutions for decision-makers to choose from each time. Furthermore, to address the poor local search capability of INSGA-II, an improvement is made by performing a neighborhood search on each new generation of the population, thereby enhancing the algorithm's local search capability and improving the quality of the solutions.

[0043] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0044] See Figures 2-5 The present invention provides a hybrid flow shop scheduling method based on INSGA-II, which includes the following steps:

[0045] Acquire processing data and establish a multi-objective optimization scheduling model. The optimization objectives in the multi-objective optimization scheduling model include completion time, energy consumption, and total load.

[0046] The initial population is generated based on a two-layer encoding / decoding rule. The gene encoding generated by the two-layer encoding / decoding rule includes a workpiece code and a machine code. The workpiece code represents the processing order of each workpiece in the first process, and the machine code represents the selection of the processing machine in each process.

[0047] Crossover and mutation operations are performed on the chromosomes in the initial population, and the crossover and / or mutation operations employ different operation methods for the work code and machine code of the chromosomes.

[0048] The chromosomes are selected, and the chromosomes that are close to the optimization target are retained as the next generation population.

[0049] A neighborhood search is performed on the next generation population to expand the next generation population.

[0050] Using the next generation population as the new initial population, the iterative update is repeated until the termination condition is met.

[0051] In some implementations, the hybrid flow shop scheduling method may further include the following steps:

[0052] After the iterative update is completed, the retained chromosomes are decoded according to the two-layer encoding / decoding rules to obtain decoding information, which is used to indicate the scheduling scheme of the hybrid flow workshop.

[0053] It should be noted that, in this invention, the order of crossover and mutation operations and selection operations on individuals is not limited to the order disclosed in the text above. In the target genetic algorithm, selection can be performed first, followed by crossover and mutation, or crossover and mutation can be performed first, followed by selection. Since the number of iterations is usually not just a few, the above order does not have a significant impact on the final iteration result. Those skilled in the art can make appropriate choices.

[0054] In some implementations, the processing data may include the number of workpieces, the number of operations, the number of machines in each operation, the processing time of each workpiece on each machine, the processing power of each machine, and the standby power.

[0055] In some implementation schemes, the completion time can be calculated as follows:

[0056]

[0057] Among them, C max Indicates the completion time; This indicates the end time of the last workpiece processed on machine k.

[0058] In some implementations, the energy consumption may include the sum of processing energy consumption and standby energy consumption, wherein the processing energy consumption is equal to the sum of the products of processing time and processing power of each machine, and the standby energy consumption is equal to the sum of the products of standby time and standby power of each machine.

[0059] In some implementations, the total load can be the sum of the total processing time for each machine, and can be calculated as follows:

[0060]

[0061] Among them, T total This represents the total load; This represents the total processing time of machine k.

[0062] As a typical application example of the above technical solution, the specific steps for establishing a multi-objective optimization scheduling model can include the following steps 1-2:

[0063] Step 1: Obtain workshop processing data. This includes the number of workpieces, the number of processes, the number of machines in each process, the processing time of each workpiece on each machine, and the processing power and standby power of each machine.

[0064] Step 2: Establish a multi-objective optimization scheduling model for the workshop.

[0065] The HFSP problem can be represented as follows: N workpieces undergo S processing steps on M machines, with multiple machines for each step, and a workpiece can be processed on any machine within that step. Constraints in the above model may include, for example:

[0066] One machine can only process one workpiece at a time.

[0067] Each workpiece can only be processed by one machine at a time.

[0068] The next process can only begin after the previous process is completed.

[0069] Once the workpiece processing begins, it cannot be interrupted.

[0070] The completion time of a workpiece in the same process consists of the start time and the processing time.

[0071] Each machine can be started at any time. The machine start time is the time when the first workpiece on the machine begins processing.

[0072] Of course, the above constraints are common constraints. In practical applications, the above constraints can be adjusted by adding or removing them according to the specific circumstances of the workshop.

[0073] A mathematical model is established with completion time, energy consumption, and total load as optimization objectives. Completion time C max The latest completion time for the machine: in This represents the end time of the last workpiece processed on machine k. Energy consumption E total Including processing energy consumption E w and standby power consumption E s E total =E w +E s Processing energy consumption equals the sum of the products of processing time and processing power for each machine, while standby energy consumption equals the sum of the products of standby time and standby power for each machine. Total load T total The sum of the total processing time for each machine: in This represents the total processing time of machine k.

[0074] Another key technical means of this invention lies in the encoding rules. For the HFSP problem, it can be divided into two sub-problems: job sequencing and machine allocation. Traditional single-layer encoding methods generally use job processing sequence codes based on job sequencing, neglecting the machine allocation problem. This is only suitable for solving single-objective optimization problems; for multi-objective optimization problems, it cannot represent the entire solution space, requiring additional heuristic rules for machine selection. Therefore, this invention adopts a two-layer encoding rule based on the first-process processing sequence code and the machine allocation code, including job codes and machine codes. The chromosome of this encoding rule consists of N+N×S gene codes. The first layer of N gene codes represents the job codes, indicating the processing order of each job in the first process; the second layer of N×S gene codes represents the machine codes, indicating the selection of processing machines in each process. Figure 3 The diagram shows an HFSP example with 4 workpieces, 2 processes, and 2 machines per process. The chromosome can be encoded as [3, 1, 4, 2; 1, 2, 2, 1; 4, 4, 3, 3]. In the first process, the workpiece processing order is 3, 1, 4, 2. Workpieces 3 and 2 are processed sequentially on machine 1, and workpieces 1 and 4 are processed sequentially on machine 2. Based on the completion time of each workpiece in the previous process, the first two completed workpieces are processed sequentially on machine 4 in the second process, according to their completion time. The last two completed workpieces are processed sequentially on machine 3, according to their completion time. This allows us to determine the workpiece processing order and the machines used in each process.

[0075] For decoding, the first part is the workpiece code. Since the processing order in subsequent processes is related to the end time of the previous process, each workpiece is processed sequentially from left to right according to the workpiece code in the first process. The processing order of workpieces in subsequent processes follows the first-come, first-served rule, that is, the workpiece that finishes in the previous process first is processed in the next process first. For the machine code, the processing machine for each workpiece in each process is determined according to the gene sequence in the chromosome code.

[0076] Once the model structure and encoding / decoding rules are determined, multiple optimal solutions can be continuously generated through iterative updates. The specific process is as follows.

[0077] In some implementations, a random initialization method can be used to generate the initial population. This invention uses a random initialization method to generate the initial population as step 3 in the application example, thereby enhancing the diversity of the initial population.

[0078] In some implementations, the crossover operation may employ a partially matching crossover operator for the workpiece code.

[0079] In some implementations, the interleaving operation can employ a two-point interleaving operator for the machine code.

[0080] In some implementations, the crossover operation may specifically include the following steps:

[0081] Two integers l1 and l2 are randomly generated in the range of 1 to N×(S+1); where N represents the number of workpieces, corresponding to the character length of the workpiece code, and S represents the number of processes, corresponding to the ratio of the character length of the machine code to N.

[0082] When l1≤N and l2≤N, the crossover operation only occurs on the workpiece code.

[0083] When l1 > N and l2 > N, the interleaving operation occurs only in the machine code.

[0084] When min(l1, l2) ≤ N and max(l1, l2) > N, the cross operation is performed on both the workpiece code and the machine code.

[0085] As a typical application example of the above technical solution, since the chromosome encoding in this invention consists of two parts, different crossover methods and rules are adopted for the workpiece code and the machine code respectively. The specific operation is as follows: Step 4, crossover operation: Two integers l1 and l2 are randomly generated in the interval 1 to N×(S+1). If l1≤N and l2≤N, the crossover operation only occurs on the workpiece code, and the partial matching crossover operator is used; if l1>N and l2>N, the crossover operation only occurs on the machine code, and the two-point crossover operator is used; if min(l1, l2)≤N and max(l1, l2)>N, the crossover operation is performed on both parts.

[0086] Then, in step 5: mutation operation, similar to the crossover operator, different mutation rules and mutation methods are used for the workpiece code and the machine code respectively. The mutation rules are the same as the crossover rules, and the mutation methods are that the workpiece code part uses the exchange mutation operator and the machine code part uses the single-point mutation operator.

[0087] In some implementations, the selection operation can be arranged at least according to the non-dominated order and crowding of the chromosomes, and can employ an elite retention strategy for selection.

[0088] In some implementations, the elite retention strategy may retain multiple chromosomes.

[0089] As a typical application example of the above technical solution, in step 6: non-dominated sorting and crowding sorting calculation, the above process performs non-dominated sorting on each chromosome individual in the population, that is, Pareto ranking. The specific operation is as follows: Assume X... a and X bThere are two feasible solutions to an n-dimensional objective minimization optimization problem, for any i = 1, 2, ..., n, we have f i (X a )≤f i (X b And there exist i = 1, 2, ..., n such that f i (X a )<f i (X b If X is the same as X, then X is called X. a DominateX b If, for a feasible solution, no other solution dominates it, then the feasible solution is called a non-dominated solution, and its Pareto level is defined as 1. Then, all currently non-dominated solutions are removed from the solution set, and the Pareto level of the remaining non-dominated solutions is defined as 2. This process continues until all solutions have a defined Pareto level. A schematic diagram of this process is shown below. Figure 4 As shown.

[0090] On the other hand, for calculating crowding degree, individuals within the same non-dominant level are ranked according to their crowding degree. The steps for calculating crowding degree are as follows:

[0091] Sort the individuals in the current non-dominated level according to the objective function, and mark the maximum value f of the objective function. max and minimum value f min .

[0092] Set the crowding of individuals at the two boundaries after sorting to infinity.

[0093] The crowding levels of the remaining individuals are calculated using the following two formulas. Where D(i) and D(i,j) represent the crowding level of individual i and the crowding level of individual i on the j-th target component, respectively. and Let be the values ​​of the preceding and following neighbors of individual i in the j-th target component. and These represent the maximum and minimum values ​​of the j-th target component, respectively.

[0094]

[0095]

[0096] Therefore, in step 7, after performing the above steps, the population is sorted according to Pareto level and crowding level, and an elite retention strategy is adopted to retain the better individuals in the population to the next generation.

[0097] Based on the above process, another key technical means of the present invention lies in performing neighborhood search on the population.

[0098] In some implementations, the neighborhood search may employ different search rules for the workpiece code and the machine code.

[0099] In some implementations, for the workpiece code, the neighborhood search can be performed using any one or a combination of two or more of the following: inverted neighborhood structure, inserted neighborhood structure, and exchanged neighborhood structure.

[0100] In some implementations, a variant neighborhood structure can be used to perform the neighborhood search for the machine code.

[0101] As typical application examples of the above technical solutions, this patent designs different search rules and neighborhood structures for the workpiece code and the machine code. For example, in step 8, the search rules in the neighborhood search are the same as the crossover and mutation rules. Because the encoding methods for the workpiece code and the machine code are different, such as... Figure 3 As shown, the genetic codes in the workpiece code are all different, while the genetic codes in the machine code for the same process may be the same. Therefore, different neighborhood structures need to be designed for the two parts of the encoding to avoid illegal solutions. For the workpiece code, three neighborhood structures are designed: reverse neighborhood structure, insertion neighborhood structure, and exchange neighborhood structure. The reverse, insertion, and exchange positions of these neighborhood structures are preferably at the positions indicated by the random numbers l1 and l2 mentioned above. The machine code uses a mutation neighborhood structure, that is, the genetic codes at the generated random numbers l1 and l2 are randomly transformed into other machine codes in the same process. The four neighborhood structures are as follows: Figure 5 As shown.

[0102] In some implementations, the termination condition includes reaching the maximum number of iterations. Specifically, for example, in step 9: determine whether the maximum number of iterations has been reached; if yes, output the optimal scheduling scheme; otherwise, execute step 4.

[0103] Based on the above technical solution, the key point of this invention is:

[0104] 1. Currently, workshop scheduling problems are mostly solved by manual scheduling based on historical experience or by using traditional scheduling algorithms. However, these methods generally optimize for a single objective. When using traditional scheduling algorithms to solve multi-objective optimization problems, they often use linear weighted summation to transform the problem into a single-objective optimization problem. This results in problems such as unreasonable weight assignment and poor objectivity. The INSGA-II proposed in this patent avoids the problem of weight assignment.

[0105] 2. To address the multi-objective scheduling problem in mixed flow workshops, a multi-objective scheduling method based on INSGA-II is proposed. Compared with the previous NSGA-II, which had poor local search capabilities, the INSGA-II proposed in this patent performs a neighborhood search on each individual in the population after each iteration of merging, which improves the local search capability of the algorithm and has better optimization effect.

[0106] As a derivative of the above method, corresponding to the hybrid flow shop scheduling method provided in any of the above embodiments, this embodiment of the invention also provides a hybrid flow shop scheduling system based on INSGA-II, which includes:

[0107] The model building module is used to acquire processing data and establish a multi-objective optimization scheduling model. The optimization objectives in the multi-objective optimization scheduling model include completion time, energy consumption, and total load.

[0108] The population generation module is used to generate an initial population based on a two-layer encoding / decoding rule. The gene encoding generated by the two-layer encoding / decoding rule includes a workpiece code and a machine code. The workpiece code represents the processing order of each workpiece in the first process, and the machine code represents the selection of the processing machine in each process.

[0109] The crossover and mutation module is used to perform crossover and mutation operations on chromosomes in the initial population. The crossover and / or mutation operations employ different operation methods for the workpiece code and machine code of the chromosomes.

[0110] The selection and retention module is used to select chromosomes and retain the chromosomes that are close to the optimization target as the next generation population.

[0111] The neighborhood search module is used to perform a neighborhood search on the next generation population in order to expand the next generation population.

[0112] The iterative update module is used to repeatedly perform iterative updates using the next generation population as the new initial population until the termination condition is met.

[0113] Accordingly, embodiments of the present invention also provide a readable storage medium storing a computer program, which, when run, executes the steps of the hybrid assembly line scheduling method provided in any of the above embodiments.

[0114] Based on the above implementation examples, it is clear that, compared with the prior art, the multi-objective scheduling method based on INSGA-II proposed in this invention avoids the problem of weight assignment in solving multi-objective optimization problems using the linear weighted summation method, and can obtain multiple non-dominated solutions for decision-makers to choose from each solution.

[0115] Furthermore, in the INSGA-II algorithm proposed in this embodiment of the invention, a neighborhood search is performed on each individual in the population after each iteration of merging, and different search rules and neighborhood structures are designed for work code and machine code respectively, thereby improving the local search capability of the algorithm and improving the quality of the solution.

[0116] In light of the above advantages, it is clear that the hybrid flow shop scheduling method provided by the embodiments of the present invention has excellent practical application effects. Specifically, for example, in practical applications, the scheduling method provided by the prior art requires manual assignment of weights to each target and then optimization, and can only provide one optimal solution at a time. In practice, the optimization results given by this method are usually quite different from the actual application requirements. For example, it often takes dozens of optimization processes to obtain results that can truly improve the shop process.

[0117] By using the implementation examples provided by this invention, satisfactory results can generally be obtained after one optimization. Based on the guidance of the results, the operational efficiency of the mixed flow workshop can be significantly improved.

[0118] It should be understood that the above embodiments are merely illustrative of the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made in accordance with the spirit and essence of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A hybrid flow shop scheduling method based on INSGA-II, characterized in that, include: Acquire processing data and establish a multi-objective optimization scheduling model, wherein the optimization objectives in the multi-objective optimization scheduling model include completion time, energy consumption, and total load; The initial population is generated based on a two-layer encoding / decoding rule. The gene encoding generated by the two-layer encoding / decoding rule includes a workpiece code and a machine code. The workpiece code represents the processing order of each workpiece in the first process, and the machine code represents the selection of the processing machine in each process. Crossover and mutation operations are performed on the chromosomes in the initial population. The crossover and / or mutation operations employ different operation methods for the work code and machine code of the chromosomes, wherein: For the workpiece code, the crossover operation uses a partial matching crossover operator; for the machine code, the crossover operation uses a two-point crossover operator. The crossover operation specifically includes: exist Two integers are randomly generated within the range. and Where N represents the number of workpieces, corresponding to the character length of the workpiece code, and S represents the number of processes, corresponding to the ratio of the character length of the machine code to N. when and When this occurs, the crossover operation only occurs on the workpiece code; when and In this case, the interleaving operation occurs only in the machine code; when and When this happens, the cross operation is performed on both the workpiece code and the machine code. Furthermore, for the workpiece code, the mutation operation employs a swap mutation operator, and for the machine code, the mutation operation employs a single-point mutation operator. A selection operation is performed on the chromosomes, retaining the portion of chromosomes that are close to the optimization target as the next generation population; A neighborhood search is performed on the next generation population to expand the next generation population, wherein: The neighborhood search employs different search rules for the workpiece code and the machine code. For the workpiece code, the neighborhood search is performed using any one or a combination of two or more of the following: reverse neighborhood structure, insertion neighborhood structure, and exchange neighborhood structure. For the machine code, the neighborhood search is performed using a mutated neighborhood structure. Using the next generation population as the new initial population, the iterative update is repeated until the termination condition is met.

2. The method for scheduling a hybrid assembly line workshop according to claim 1, characterized in that, Also includes: After the iterative update is completed, the retained chromosomes are decoded according to the two-layer encoding / decoding rules to obtain decoding information, which is used to indicate the scheduling scheme of the hybrid flow workshop.

3. The method for scheduling a hybrid assembly line workshop according to claim 1, characterized in that, The processing data includes the number of workpieces, the number of processes, the number of machines in each process, the processing time of each workpiece on each machine, the processing power of each machine, and the standby power.

4. The method for scheduling a hybrid assembly line workshop according to claim 3, characterized in that, The completion time is calculated as follows: ； in, Indicates the completion time; Indicates machine The end time of the last processed workpiece; And / or, the energy consumption includes the sum of processing energy consumption and standby energy consumption, wherein the processing energy consumption is equal to the sum of the products of processing time and processing power of each machine, and the standby energy consumption is equal to the sum of the products of standby time and standby power of each machine. And / or, the total load is the sum of the total processing times of each machine, calculated as follows: ； in, This represents the total load; Indicates machine Total processing time.

5. The method for scheduling a hybrid assembly line workshop according to claim 1, characterized in that, The initial population is generated using a random initialization method.

6. The method for scheduling a hybrid assembly line workshop according to claim 1, characterized in that, The selection operation is performed by arranging the chromosomes at least according to their non-dominated order and crowding calculations, and by employing an elite retention strategy.

7. The method for scheduling a hybrid assembly line workshop according to claim 6, characterized in that, The elite preservation strategy preserves multiple chromosomes.

8. The method for scheduling a hybrid assembly line workshop according to claim 1, characterized in that, The termination condition includes reaching the maximum number of iterations.

9. A hybrid flow shop scheduling system based on INSGA-II for implementing the hybrid flow shop scheduling method according to any one of claims 1-8, characterized in that, include: The model building module is used to acquire processing data and build a multi-objective optimization scheduling model. The optimization objectives in the multi-objective optimization scheduling model include completion time, energy consumption, and total load. The population generation module is used to generate an initial population based on a two-layer encoding / decoding rule. The gene encoding generated by the two-layer encoding / decoding rule includes a workpiece code and a machine code. The workpiece code represents the processing order of each workpiece in the first process, and the machine code represents the selection of the processing machine in each process. The crossover and mutation module is used to perform crossover and mutation operations on chromosomes in the initial population. The crossover and / or mutation operations employ different operation methods for the work code and machine code of the chromosomes. The selection and retention module is used to select chromosomes and retain a portion of chromosomes that are close to the optimization target as the next generation population. The neighborhood search module is used to perform a neighborhood search on the next generation population in order to expand the next generation population; The iterative update module is used to repeatedly perform iterative updates using the next generation population as the new initial population until the termination condition is met.

10. A readable storage medium, characterized in that, The readable storage medium stores a computer program, which, when run, executes the steps of the hybrid flow shop scheduling method according to any one of claims 1-8.

Images