A product assembly scheduling method, device and equipment

Through integer programming model and double-layer hybrid coding, the workers' allocation and task order are optimized, and the improved multi-objective evolution algorithm (MOEA) is used to solve the problems of low production efficiency and uneven busyness in the assembly process of complex products, achieving more efficient Production and resource utilization.

CN114677028BActive Publication Date: 2024-11-01BEIJING INST OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210345464.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-03-31
Publication Date
2024-11-01
Estimated Expiration
2042-03-31

AI Technical Summary

Technical Problem

The problems of low production efficiency and unbalanced work of assembly workers during complex product assembly lead to low production efficiency and waste of human resources.

Method used

An integer programming model is used to combine two-layer hybrid coding and an improved multi-objective evolution algorithm (MOEA) to optimize worker allocation and task order to minimize maximum completion time and busy and idle imbalance and improve production efficiency.

Benefits of technology

By optimizing worker allocation and task arrangement, production efficiency is significantly improved, the uneven busyness of assembly workers is reduced, and the efficiency of human resources is improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114677028B_ABST
    Figure CN114677028B_ABST
Patent Text Reader

Abstract

The present invention provides a product assembly scheduling method, device and equipment, the product assembly scheduling method comprising: establishing an integer programming model according to product assembly information; determining a worker allocation code and a task sequence code according to the product assembly information and the integer programming model using a double-layer mixed coding method; obtaining multiple product assembly scheduling schemes based on the integer programming model and the worker allocation code and the task sequence code; and determining a product assembly scheduling optimization scheme from the multiple product assembly scheduling schemes using an improved multi-objective improved algorithm MOEA. The scheme of the present invention can improve production efficiency and reduce the occurrence of uneven busy and idle assembly workers.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of product assembly, and in particular to a product assembly scheduling method, device and equipment. Background Art

[0002] Complex products are products with complex requirements, complex components, complex technologies, complex processes, and complex management, such as satellites, missiles, and aircraft. Assembly is the last link in the development and production of complex products and plays an important role. According to statistics, assembly accounts for nearly 50% of the total production time of products, 20% of the total production cost, and 30% to 50% of labor costs. Aerospace products such as satellites and missiles are typical complex products, and their assembly process has the characteristics of order-oriented production, mainly manual assembly, and wide range of professional fields. To fully realize the assembly process of aerospace products, not only a large amount of human resources is required, but also a wide variety of materials, workers, tooling and other assembly resources are involved. In addition, parallel teams are set up on critical process paths. Therefore, production efficiency is often not high and scheduling is difficult. At present, the main scheduling methods are often based on expert experience or heuristic rules, which are too subjective, poorly stable, and low in scheduling efficiency. In addition, unreasonable scheduling planning can easily cause uneven workload of workers and waste of human resources. In the case of manual work and machine work, workers are more likely to get fatigued, resulting in reduced work efficiency. Therefore, coordinating the workload among workers is also an issue that cannot be ignored. Summary of the invention

[0003] The embodiments of the present invention provide a product assembly scheduling method, device and equipment to solve the problems of low production efficiency and unbalanced work schedule of assembly workers in the existing product assembly process.

[0004] In order to solve the above technical problems, the embodiments of the present invention provide the following technical solutions:

[0005] An embodiment of the present invention provides a product assembly scheduling method, comprising:

[0006] An integer programming model is established according to the product assembly information; the integer programming model includes a first objective function for indicating the maximum value of minimizing the completion time of the team, a second objective function for indicating minimizing the busy-idle imbalance, and a constraint condition for indicating the assembly workers in the team to perform assembly work;

[0007] According to the product assembly information and the integer programming model, a worker allocation code and a task sequence code are determined by using a double-layer mixed coding method; the worker allocation code is used to indicate the work position that the assembly worker can go to; the task sequence code is used to indicate the order in which the assembly tasks enter the assembly island;

[0008] Based on the integer programming model, a plurality of product assembly scheduling schemes are obtained according to the worker allocation code and the task sequence code;

[0009] By utilizing the improved multi-objective improved algorithm MOEA, the product assembly scheduling optimization scheme is determined among the multiple product assembly scheduling schemes.

[0010] Optionally, the product assembly information includes:

[0011] Assembly island area information;

[0012] Team information in the assembly island;

[0013] All the skills information required for assembly tasks in the assembly island;

[0014] Assembly worker information;

[0015] Assembly task information;

[0016] The coordinate information of the work positions in the team to which the assembly workers can go;

[0017] the skills that assemblers possess and the proficiency of the skills they possess;

[0018] In an assembly task, the basic working hours required for an assembly worker to complete the assembly task corresponding to his or her skills.

[0019] Optionally, the second objective function is determined based on the mean square error of the maximum working time period of each team and the mean square error of the total working time period between teams.

[0020] Optionally, the constraint conditions include:

[0021] The first constraint is used to indicate the skills that the assembly workers have and the proficiency of the skills they have;

[0022] The second constraint is used to indicate whether the assembly worker works in a team;

[0023] The third constraint is used to indicate whether the assembly task is processed on the team;

[0024] The fourth constraint is used to indicate whether the team needs to handle the work tasks corresponding to the skill;

[0025] The fifth constraint is used to indicate the processing order of assembly tasks on the team;

[0026] The sixth constraint is used to indicate the degree to which the proficiency of the assembly workers' skills affects the working hours;

[0027] The seventh constraint is used to indicate the working time of the team in processing the assembly task;

[0028] The eighth constraint condition is used to indicate that the assembly worker can be assigned to a team that requires the target skill to handle the corresponding work task only if the proficiency of the target skill possessed by the assembly worker is not 0;

[0029] The ninth constraint is used to indicate the number of assembly workers assigned to all teams;

[0030] The tenth constraint is used to indicate that the same team on the same assembly island must have the same skills;

[0031] The eleventh constraint is used to indicate the different skills that different teams on the same assembly island must possess;

[0032] The twelfth constraint is used to indicate that each assembly task can only be assigned to one team on each assembly island;

[0033] The thirteenth constraint is used to indicate that the start time of the assembly task on the team is 0, or the end time of the previous assembly task processed by the team;

[0034] The fourteenth constraint condition is used to indicate that the end time of the assembly task on the team is the sum of the start time of the assembly task on the team and the processing time of the assembly task on the team;

[0035] The fifteenth constraint is used to indicate that an assembly task cannot be processed on multiple assembly islands at the same time;

[0036] The sixteenth constraint is used to indicate that a team can only handle one assembly task at the same time.

[0037] Optionally, a double-layer hybrid coding method is used to determine the worker allocation code and the task sequence code, including:

[0038] Determine the worker allocation code by integer coding;

[0039] The task sequence code is determined by using an arrangement coding method.

[0040] Optionally, according to the product assembly information and the integer programming model, a worker allocation code is determined using a double-layer hybrid coding method, including:

[0041] Obtaining required position information according to all skill information required for the assembly task in the assembly island in the product assembly information and the team information in the assembly island in the product assembly information;

[0042] The worker allocation code is determined according to the required job information and the skills of the assembly workers while satisfying the constraint conditions.

[0043] Optionally, based on the integer programming model, multiple product assembly scheduling solutions are obtained according to the worker allocation code and the task sequence code, including:

[0044] Determining first worker information according to the worker allocation code; the first worker information includes allocation information of assembly workers working on the team;

[0045] Determine first working time information according to the first worker information; the first working time information includes the working time of the assembly workers in each team when processing the assembly task;

[0046] Determine task sequence information according to the task sequence code and the first working time information; the task sequence information includes the order in which each assembly task enters each assembly island;

[0047] Determine the first team information according to the first working hour information and the team work information; the first team information is used to indicate that the target assembly task is processed on the target team; the target assembly task is one of each of the assembly tasks; the target team is one of the teams; the team work information includes: the start time of the team processing the current assembly task, the end time of the team processing the current assembly task and the working time of the team processing the assembly task;

[0048] The plurality of product assembly scheduling plans are obtained according to the task sequence information, the first team information and the first worker information.

[0049] Optionally, the method further comprises:

[0050] Determine a first function value of the first objective function corresponding to each of the product assembly scheduling plans;

[0051] Determine a second function value of the second objective function corresponding to each of the product assembly scheduling plans.

[0052] Optionally, determining first team information according to the first working hour information and team work information includes:

[0053] Determine, among each of the teams, the first team that has the closest end time for processing the current assembly task as the target team;

[0054] In the case where there are multiple first shifts, among the first shifts, the second shift with the shortest working time is determined as the target shift;

[0055] In the case that there are multiple second teams, the third team in the second teams is determined as the target team; the third team is any one of the second teams.

[0056] Optionally, according to the improved multi-objective improved algorithm MOEA, before the multiple product assembly scheduling schemes determine the product assembly scheduling optimization scheme, the method further includes:

[0057] Through the improvement strategy, the multi-objective improved algorithm MOEA is improved to obtain the improved multi-objective improved algorithm MOEA;

[0058] The improvement strategy includes at least one of the following:

[0059] Improved search mechanism strategies;

[0060] Improved strategies for evolving population structure;

[0061] Improve the population distribution maintenance mechanism strategy.

[0062] Optionally, the multi-objective improved algorithm MOEA is improved by improving the search mechanism strategy to obtain an improved multi-objective improved algorithm MOEA, including:

[0063] According to a preset mutation mode suitable for permutation coding and a preset mutation mode suitable for integer coding, a mutation combination complex scheme is obtained;

[0064] According to the mutation combination scheme, the evolved offspring are further searched to obtain the mutant offspring, so that the improved multi-objective improved algorithm MOEA searches among the evolved offspring, the mutant offspring and the population parent generation.

[0065] Optionally, the multi-objective improved algorithm MOEA is improved by improving the evolutionary population structure strategy to obtain an improved multi-objective improved algorithm MOEA, including:

[0066] Divide the evolutionary offspring into non-dominant hierarchies;

[0067] Determine the evolutionary offspring whose non-dominant level is less than or equal to the first preset level as the elite offspring;

[0068] The evolutionary offspring are searched according to the elite offspring to obtain the mutant elite offspring, so that the improved multi-objective improved algorithm MOEA searches in the mutant elite offspring.

[0069] Optionally, the multi-objective improved algorithm MOEA is improved by improving the population distribution maintaining mechanism strategy to obtain an improved multi-objective improved algorithm MOEA, including:

[0070] Divide the evolutionary offspring into non-dominant hierarchies;

[0071] According to the non-kinship calculation function, the crowding distance strategy is improved so that the improved multi-objective improved algorithm MOEA generates the next generation of individuals according to the evolutionary offspring of the target non-dominated hierarchy;

[0072] Wherein, the non-kinship calculation function is determined based on the crowding distance and the similarity;

[0073] The target non-dominated hierarchy is a non-dominated hierarchy in which the total number of individuals is less than the preset number of individuals in the population, and the total number of individuals after merging is greater than the preset number of individuals in the population.

[0074] Optionally, the method further comprises:

[0075] Analyze the complexity of the improved first multi-objective improved algorithm MOEA to obtain a first complexity analysis result; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved search mechanism strategy;

[0076] Analyze the complexity of the improved second multi-objective improved algorithm MOEA to obtain a second complexity analysis result; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved evolutionary population structure strategy;

[0077] Analyze the complexity of the improved third multi-objective improved algorithm MOEA to obtain a third complexity analysis result; the improved third multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved population distribution maintaining mechanism strategy;

[0078] Analyze the complexity of the preset multi-objective improved algorithm to obtain a fourth complexity analysis result;

[0079] The first complexity analysis result, the second complexity analysis result, the third complexity analysis result and the fourth complexity analysis result are compared to obtain a first comparison result indicating the complexity of the improved multi-objective improved algorithm MOEA.

[0080] Optionally, the method further comprises:

[0081] Obtaining a first assembly scheduling test plan according to a preset number of first assembly islands, a preset number of first assembly tasks, and a preset number of first assembly workers;

[0082] By using the improved first multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, a first union of non-dominated solutions and a first Pareto front of non-dominated solutions in the first union are obtained, wherein the first Pareto front is used to indicate the ability of the improved first multi-objective optimization algorithm MOEA to approach the optimal solution; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved search mechanism strategy;

[0083] By using the improved second multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, a second union of non-dominated solutions and a second Pareto front of non-dominated solutions in the second union are obtained, wherein the second Pareto front is used to indicate the ability of the improved second multi-objective optimization algorithm MOEA to approach the optimal solution; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0084] By using the improved fourth multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, a third union of non-dominated solutions and a third Pareto front of non-dominated solutions in the third union are obtained, wherein the third Pareto front is used to indicate the ability of the improved fourth multi-objective optimization algorithm MOEA to approach the optimal solution; the improved fourth multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0085] Obtaining, by using a preset multi-objective optimization algorithm and the first assembly scheduling test scheme, a fourth union of non-dominated solutions and a fourth Pareto front of the non-dominated solutions in the fourth union, wherein the fourth Pareto front is used to indicate the ability of the preset multi-objective optimization algorithm to approach an optimal solution;

[0086] The first Pareto front, the second Pareto front, the third Pareto front and the fourth Pareto front are compared to obtain a second comparison result indicating the ability of the improved multi-objective optimization algorithm MOEA to approach the optimal solution.

[0087] Optionally, the method further comprises:

[0088] Obtaining a second assembly scheduling test plan according to a preset number of second assembly islands, a preset number of second assembly tasks, and a preset number of second assembly workers;

[0089] By using the improved first multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, a first union of non-dominated solutions is obtained, and according to the first union, a first inverse generation distance analysis result and a first hypervolume index analysis result of the improved first multi-objective optimization algorithm MOEA are obtained; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved search mechanism strategy;

[0090] By using the improved second multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, a second union of non-dominated solutions is obtained, and according to the second union, a second inverse generation distance analysis result and a second hypervolume index analysis result of the improved second multi-objective optimization algorithm MOEA are obtained; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0091] By using the improved fourth multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, a third union of non-dominated solutions is obtained, and according to the third union, a third inverse generation distance analysis result and a third hypervolume index analysis result of the improved fourth multi-objective optimization algorithm MOEA are obtained; the improved fourth multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0092] Obtaining a fourth union of non-dominated solutions through a preset multi-objective optimization algorithm and the second assembly scheduling test scheme, and obtaining a fourth inverse generation distance analysis result and a fourth hypervolume index analysis result of the preset multi-objective optimization algorithm according to the third union;

[0093] Comparing the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result and the fourth inverse generation distance analysis result to obtain a third comparison result indicating the convergence of the improved multi-objective optimization algorithm MOEA;

[0094] The first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result are compared to obtain a fourth comparison result indicating the convergence of the improved multi-objective optimization algorithm MOEA.

[0095] Optionally, the method further comprises:

[0096] Obtaining a fifth comparison result indicating the stability of the improved multi-objective optimization algorithm MOEA according to the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result, and the fourth inverse generation distance analysis result;

[0097] According to the first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result, a sixth comparison result indicating the stability of the improved multi-objective optimization algorithm MOEA is obtained.

[0098] Optionally, the method further comprises:

[0099] Taking the improved first multi-objective optimization algorithm MOEA as a benchmark, obtaining a seventh comparison result indicating the dominance of the improved multi-objective optimization algorithm MOEA according to the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result and the fourth inverse generation distance analysis result;

[0100] Taking the improved first multi-objective optimization algorithm MOEA as a benchmark, according to the first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result, an eighth comparison result indicating the dominance of the improved multi-objective optimization algorithm MOEA is obtained.

[0101] Optionally, the method further comprises:

[0102] According to the first Pareto front, the second Pareto front and the third Pareto front, a ninth comparison result indicating the applicability of the improved multi-objective optimization algorithm MOEA is obtained.

[0103] Optionally, the method further comprises:

[0104] According to the first inverse generation distance analysis result, the second inverse generation distance analysis result and the third inverse generation distance analysis result, a tenth comparison result indicating the applicability of the improved multi-objective optimization algorithm MOEA is obtained.

[0105] Optionally, the method further comprises:

[0106] According to the first hypervolume index analysis result, the second hypervolume index analysis result and the third hypervolume index analysis result, an eleventh comparison result is obtained for indicating the applicability of the improved multi-objective optimization algorithm MOEA.

[0107] The embodiment of the present invention further provides a product assembly scheduling device, comprising:

[0108] A model building module, used to build an integer programming model according to product assembly information; the integer programming model includes a first objective function for indicating the maximum value of minimizing the completion time of the team, a second objective function for indicating minimizing the busy-idle imbalance, and a constraint condition for indicating the assembly workers in the team to perform assembly work;

[0109] A first determination module is used to determine a worker allocation code and a task sequence code according to the product assembly information and the integer programming model by using a double-layer mixed coding method; the worker allocation code is used to indicate the work position that the assembly worker can go to; the task sequence code is used to indicate the order in which the assembly tasks enter the assembly island;

[0110] A second determination module is used to obtain a plurality of product assembly scheduling solutions based on the integer programming model and according to the worker allocation code and the task sequence code;

[0111] The third determination module is used to determine the product assembly scheduling optimization solution from the multiple product assembly scheduling solutions by using the improved multi-objective improved algorithm MOEA.

[0112] Optionally, the product assembly information includes:

[0113] Assembly island area information;

[0114] Team information in the assembly island;

[0115] All the skills information required for assembly tasks in the assembly island;

[0116] Assembly worker information;

[0117] Assembly task information;

[0118] The coordinate information of the work positions in the team to which the assembly workers can go;

[0119] the skills that assemblers possess and the proficiency of the skills they possess;

[0120] In an assembly task, the basic working hours required for an assembly worker to complete the assembly task corresponding to his or her skills.

[0121] Optionally, the second objective function is determined based on the mean square error of the maximum working time period of each team and the mean square error of the total working time period between teams.

[0122] Optionally, the constraint conditions include:

[0123] The first constraint is used to indicate the skills that the assembly workers have and the proficiency of the skills they have;

[0124] The second constraint is used to indicate whether the assembly worker works in a team;

[0125] The third constraint is used to indicate whether the assembly task is processed on the team;

[0126] The fourth constraint is used to indicate whether the team needs to handle the work tasks corresponding to the skill;

[0127] The fifth constraint is used to indicate the processing order of assembly tasks on the team;

[0128] The sixth constraint is used to indicate the degree to which the proficiency of the assembly workers' skills affects the working hours;

[0129] The seventh constraint is used to indicate the working time of the team in processing the assembly task;

[0130] The eighth constraint condition is used to indicate that the assembly worker can be assigned to a team that requires the target skill to handle the corresponding work task only if the proficiency of the target skill possessed by the assembly worker is not 0;

[0131] The ninth constraint is used to indicate the number of assembly workers assigned to all teams;

[0132] The tenth constraint is used to indicate that the same team on the same assembly island must have the same skills;

[0133] The eleventh constraint is used to indicate the different skills that different teams on the same assembly island must possess;

[0134] The twelfth constraint is used to indicate that each assembly task can only be assigned to one team on each assembly island;

[0135] The thirteenth constraint is used to indicate that the start time of the assembly task on the team is 0, or the end time of the previous assembly task processed by the team;

[0136] The fourteenth constraint condition is used to indicate that the end time of the assembly task on the team is the sum of the start time of the assembly task on the team and the processing time of the assembly task on the team;

[0137] The fifteenth constraint is used to indicate that an assembly task cannot be processed on multiple assembly islands at the same time;

[0138] The sixteenth constraint is used to indicate that a team can only handle one assembly task at the same time.

[0139] Optionally, the first determining module includes:

[0140] A first determining unit, configured to determine the worker allocation code by using integer coding;

[0141] The second determining unit is used to determine the task sequence code by using an arrangement coding method.

[0142] Optionally, the first determining module includes:

[0143] A third determining unit is used to obtain required position information according to all skill information required for the assembly task in the assembly island in the product assembly information and the team information in the assembly island in the product assembly information;

[0144] The fourth determining unit is used to determine the worker allocation code according to the required job information and the skills of the assembly worker while satisfying the constraint condition.

[0145] Optionally, the second determining module includes:

[0146] a fifth determining unit, configured to determine first worker information according to the worker allocation code; the first worker information includes allocation information of assembly workers working on the team;

[0147] a sixth determining unit, configured to determine first working time information according to the first worker information; the first working time information includes working hours of assembly workers in each team when processing assembly tasks;

[0148] a seventh determining unit, configured to determine task sequence information according to the task sequence code and the first working time information; the task sequence information includes the order in which each assembly task enters each assembly island;

[0149] an eighth determination unit, configured to determine first team information according to the first working hour information and team work information; the first team information is used to indicate that a target assembly task is processed on a target team; the target assembly task is one of each of the assembly tasks; the target team is one of the teams; the team work information includes: a start time when the team processes the current assembly task, an end time when the team processes the current assembly task, and a working time for the team to process the assembly task;

[0150] A ninth determining unit is used to obtain the plurality of product assembly scheduling plans according to the task sequence information, the first team information and the first worker information.

[0151] Optionally, the second determining module further includes:

[0152] A tenth determining unit, configured to determine a first function value of the first objective function corresponding to each of the product assembly scheduling plans;

[0153] An eleventh determining unit is used to determine a second function value of the second objective function corresponding to each of the product assembly scheduling plans.

[0154] Optionally, the eighth determining unit is specifically configured to:

[0155] Determine, among each of the teams, the first team that has the closest end time for processing the current assembly task as the target team;

[0156] In the case where there are multiple first shifts, among the first shifts, the second shift with the shortest working time is determined as the target shift;

[0157] In the case that there are multiple second teams, the third team in the second teams is determined as the target team; the third team is any one of the second teams.

[0158] Optionally, the device further comprises:

[0159] An algorithm improvement module is used to improve the multi-objective improved algorithm MOEA through an improvement strategy to obtain an improved multi-objective improved algorithm MOEA;

[0160] The improvement strategy includes at least one of the following:

[0161] Improved search mechanism strategies;

[0162] Improved strategies for evolving population structure;

[0163] Improve the population distribution maintenance mechanism strategy.

[0164] Optionally, the algorithm improvement module includes:

[0165] A first processing unit, configured to obtain a mutation combination complex scheme according to a preset mutation mode suitable for permutation coding and a preset mutation mode suitable for integer coding;

[0166] The first search unit is used to further search the evolved offspring according to the mutation combination scheme to obtain the mutant offspring, so that the improved multi-objective improved algorithm MOEA searches among the evolved offspring, the mutant offspring and the population parent generation.

[0167] Optionally, the algorithm improvement module includes:

[0168] The first level division unit is used to divide the evolutionary offspring into non-dominated levels;

[0169] A twelfth determination unit, used to determine the evolutionary offspring whose non-dominant level is less than or equal to the first preset level as the elite offspring;

[0170] The second search unit is used to search the evolutionary offspring according to the elite offspring to obtain the mutant elite offspring, so that the improved multi-objective improved algorithm MOEA searches in the mutant elite offspring.

[0171] Optionally, the algorithm improvement module includes:

[0172] The second level division unit is used to divide the evolutionary offspring into non-dominated levels;

[0173] The second processing unit is used to improve the crowding distance strategy according to the non-close relative degree calculation function, so that the improved multi-objective improved algorithm MOEA generates the next generation of individuals according to the evolutionary offspring of the target non-dominated level;

[0174] Wherein, the non-kinship calculation function is determined based on the crowding distance and the similarity;

[0175] The target non-dominated hierarchy is a non-dominated hierarchy in which the total number of individuals is less than the preset number of individuals in the population, and the total number of individuals after merging is greater than the preset number of individuals in the population.

[0176] Optionally, the algorithm improvement module further includes:

[0177] A first analysis unit is used to analyze the complexity of the improved first multi-objective improved algorithm MOEA to obtain a first complexity analysis result; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved search mechanism strategy;

[0178] The second analysis unit is used to analyze the complexity of the improved second multi-objective improved algorithm MOEA to obtain a second complexity analysis result; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved evolutionary population structure strategy;

[0179] The third analysis unit is used to analyze the complexity of the improved third multi-objective improved algorithm MOEA to obtain a third complexity analysis result; the improved third multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved population distribution maintaining mechanism strategy;

[0180] A fourth analysis unit, used to analyze the complexity of the preset multi-objective improved algorithm to obtain a fourth complexity analysis result;

[0181] The first comparison unit is used to compare the first complexity analysis result, the second complexity analysis result, the third complexity analysis result and the fourth complexity analysis result to obtain a first comparison result indicating the complexity of the improved multi-objective improved algorithm MOEA.

[0182] Optionally, the algorithm improvement module further includes:

[0183] A thirteenth determining unit, configured to obtain a first assembly scheduling test plan according to a preset number of first assembly islands, a preset number of first assembly tasks, and a preset number of first assembly workers;

[0184] A third processing unit is used to obtain a first union of non-dominated solutions and a first Pareto front of non-dominated solutions in the first union through an improved first multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, wherein the first Pareto front is used to indicate the ability of the improved first multi-objective optimization algorithm MOEA to approach an optimal solution; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved search mechanism strategy;

[0185] a fourth processing unit, configured to obtain a second union of non-dominated solutions and a second Pareto front of non-dominated solutions in the second union by using an improved second multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, wherein the second Pareto front is used to indicate the ability of the improved second multi-objective optimization algorithm MOEA to approach an optimal solution; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0186] a fifth processing unit, configured to obtain a third union of non-dominated solutions and a third Pareto front of non-dominated solutions in the third union by using an improved fourth multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, wherein the third Pareto front is used to indicate the ability of the improved fourth multi-objective optimization algorithm MOEA to approach an optimal solution; the improved fourth multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0187] a sixth processing unit, configured to obtain, by using a preset multi-objective optimization algorithm and the first assembly scheduling test scheme, a fourth union of non-dominated solutions and a fourth Pareto front of the non-dominated solutions in the fourth union, wherein the fourth Pareto front is used to indicate an ability of the preset multi-objective optimization algorithm to approach an optimal solution;

[0188] The second comparison unit is used to compare the first Pareto front, the second Pareto front, the third Pareto front and the fourth Pareto front to obtain a second comparison result indicating the ability of the improved multi-objective optimization algorithm MOEA to approach the optimal solution.

[0189] Optionally, the algorithm improvement module further includes:

[0190] A fourteenth determining unit, configured to obtain a second assembly scheduling test plan according to a preset number of second assembly islands, a preset number of second assembly tasks, and a preset number of second assembly workers;

[0191] a seventh processing unit, configured to obtain a first union of non-dominated solutions through the improved first multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, and obtain a first inverse generation distance analysis result and a first hypervolume index analysis result of the improved first multi-objective optimization algorithm MOEA according to the first union; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved search mechanism strategy;

[0192] an eighth processing unit, configured to obtain a second union of non-dominated solutions through an improved second multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, and to obtain a second inverse generation distance analysis result and a second hypervolume index analysis result of the improved second multi-objective optimization algorithm MOEA according to the second union; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved evolutionary population structure strategy;

[0193] a ninth processing unit, configured to obtain a third union of non-dominated solutions through an improved fourth multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, and to obtain a third inverse generation distance analysis result and a third hypervolume index analysis result of the improved fourth multi-objective optimization algorithm MOEA according to the third union; the improved fourth multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved evolutionary population structure strategy;

[0194] a tenth processing unit, configured to obtain a fourth union of non-dominated solutions by using a preset multi-objective optimization algorithm and the second assembly scheduling test scheme, and to obtain a fourth inverse generation distance analysis result and a fourth hypervolume index analysis result of the preset multi-objective optimization algorithm according to the third union;

[0195] A third comparison unit is used to compare the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result and the fourth inverse generation distance analysis result to obtain a third comparison result indicating the convergence of the improved multi-objective optimization algorithm MOEA;

[0196] The fourth comparison unit is used to compare the first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result to obtain a fourth comparison result indicating the convergence of the improved multi-objective optimization algorithm MOEA.

[0197] Optionally, the algorithm improvement module further includes:

[0198] an eleventh processing unit, configured to obtain a fifth comparison result indicating the stability of the improved multi-objective optimization algorithm MOEA according to the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result, and the fourth inverse generation distance analysis result;

[0199] The twelfth processing unit is used to obtain a sixth comparison result indicating the stability of the improved multi-objective optimization algorithm MOEA based on the first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result.

[0200] Optionally, the algorithm improvement module further includes:

[0201] A thirteenth processing unit is used to obtain a seventh comparison result indicating the dominance of the improved multi-objective optimization algorithm MOEA based on the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result and the fourth inverse generation distance analysis result, taking the improved first multi-objective optimization algorithm MOEA as a benchmark;

[0202] The fourteenth processing unit is used to obtain an eighth comparison result indicating the dominance of the improved multi-objective optimization algorithm MOEA based on the first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result, taking the improved first multi-objective optimization algorithm MOEA as a benchmark.

[0203] Optionally, the algorithm improvement module further includes:

[0204] The fifteenth processing unit is used to obtain a ninth comparison result indicating the applicability of the improved multi-objective optimization algorithm MOEA based on the first Pareto front, the second Pareto front and the third Pareto front.

[0205] Optionally, the algorithm improvement module further includes:

[0206] The sixteenth processing unit is used to obtain a tenth comparison result indicating the applicability of the improved multi-objective optimization algorithm MOEA based on the first inverse generation distance analysis result, the second inverse generation distance analysis result and the third inverse generation distance analysis result.

[0207] Optionally, the algorithm improvement module further includes:

[0208] The seventeenth processing unit is used to obtain an eleventh comparison result indicating the applicability of the improved multi-objective optimization algorithm MOEA based on the first hypervolume index analysis result, the second hypervolume index analysis result and the third hypervolume index analysis result.

[0209] An embodiment of the present invention also provides a product assembly scheduling device, comprising: a processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the steps of the product assembly scheduling method as described in any one of the above items are implemented.

[0210] An embodiment of the present invention further provides a readable storage medium, on which a program is stored. When the program is executed by a processor, the steps in the product assembly scheduling method as described above are implemented.

[0211] The beneficial effects of the present invention are:

[0212] The scheme of the present invention establishes, based on product assembly information, an integer programming model including a first objective function for indicating the maximum value of minimizing the team completion time, a second objective function for indicating minimizing the busy-idle imbalance, and an integer programming model for indicating the constraints for assembly workers in the team to perform assembly work, with minimizing the maximum completion time and minimizing the busy-idle imbalance as optimization goals, and through hybrid encoding and decoding, multiple product assembly scheduling schemes and an improved multi-objective improved algorithm MOEA are obtained to obtain a product assembly scheduling optimization scheme, which can improve production efficiency and reduce the occurrence of busy-idle imbalance among assembly workers. BRIEF DESCRIPTION OF THE DRAWINGS

[0213] Figure 1 A flowchart showing a product assembly scheduling method provided by an embodiment of the present invention;

[0214] Figure 2A schematic diagram showing double-layer hybrid coding provided by an embodiment of the present invention;

[0215] Figure 3 A schematic diagram showing a method for determining a worker allocation code according to an embodiment of the present invention;

[0216] Figure 4 A schematic diagram showing an improved strategy 1 provided in an embodiment of the present invention;

[0217] Figure 5 A schematic diagram showing an improved strategy 2 provided in an embodiment of the present invention;

[0218] Figure 6 A schematic diagram showing an improved strategy 3 provided in an embodiment of the present invention;

[0219] Figure 7 A schematic diagram showing the distribution of skills provided by an embodiment of the present invention on an assembly island;

[0220] Figure 8 represents one of the PF scatter plots provided by an embodiment of the present invention;

[0221] Fig. 9 It represents the second PF scatter diagram provided by the embodiment of the present invention;

[0222] Fig.10 The third PF scatter diagram provided by the embodiment of the present invention is shown;

[0223] Fig.11 The fourth PF scatter diagram provided by the embodiment of the present invention is shown;

[0224] Fig.12 It represents one of the IGD indicator trend graphs provided by the embodiment of the present invention;

[0225] Fig.13 The second IGD indicator trend diagram provided by the embodiment of the present invention is shown;

[0226] Fig.14 The third IGD indicator trend diagram provided by the embodiment of the present invention is shown;

[0227] Fig.15 The fourth IGD indicator trend diagram provided by the embodiment of the present invention is shown;

[0228] Fig.16 It represents one of the HV index trend graphs provided by the embodiment of the present invention;

[0229] Fig.17 The second HV index trend diagram provided by the embodiment of the present invention is shown;

[0230] Fig.18 The third HV index trend diagram provided by the embodiment of the present invention is shown;

[0231] Fig.19 The fourth HV index trend diagram provided by the embodiment of the present invention is shown;

[0232] Fig. 20 One of the box plots representing the IGD indicator provided by the embodiment of the present invention;

[0233] Fig.21 The second box plot showing the IGD indicator provided by the embodiment of the present invention;

[0234] Fig. 22 A third box plot showing the IGD indicator provided by an embodiment of the present invention;

[0235] Fig.23 A fourth box plot showing the IGD indicator provided by an embodiment of the present invention;

[0236] Fig.24 One of the box plots representing the HV index provided by the embodiment of the present invention;

[0237] Fig.25 The second box plot showing the HV index provided by the embodiment of the present invention;

[0238] Fig.26 A third box plot showing the HV index provided by an embodiment of the present invention;

[0239] Fig. 27 A fourth box plot showing the HV index provided by an embodiment of the present invention;

[0240] Fig.28 A schematic diagram showing the structure of a product assembly scheduling device provided by an embodiment of the present invention;

[0241] Fig.29 A schematic diagram showing the structure of a product assembly scheduling device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0242] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0243] Aiming at the problems of low production efficiency and unbalanced work schedule of assembly workers in the existing product assembly process, the present invention provides a product assembly scheduling method, device and equipment.

[0244] like Figure 1 As shown, an embodiment of the present invention provides a product assembly scheduling method, comprising:

[0245] Step 101: Establish an integer programming model based on product assembly information; the integer programming model includes a first objective function for indicating the maximum value of minimizing the team completion time, a second objective function for indicating minimizing the busy-idle imbalance, and constraints for indicating assembly workers in the team to perform assembly work.

[0246] It should be noted that the embodiment of the present invention uses the complex product assembly line scheduling problem, refines its process critical path as a hybrid assembly line model, and considers the allocation problem of multi-skilled and multi-level assembly workers on the assembly line, and proposes a hybrid assembly line worker allocation problem (Workers Assignment Problem of Hybrid Assembly Line, WAPHAL). The difficulty of this problem lies in that the parallel machines in the hybrid assembly line are unrelated, and the teams are not machines but assembly workers. In addition, the allocation of assembly workers must not only consider the quantitative allocation, but also the skills and levels of the workers. It is a multi-variable coupled non-deterministic polynomial (NP) hard problem.

[0247] There are a large number of parallel processing processes in the assembly tasks of small batches of aerospace products, so parallel teams are set up during the assembly process to improve production efficiency. At the same time, as a complex product, aerospace products not only rely on manual assembly dominated by people, but also involve skills and trades with large differences and large numbers. Based on these characteristics, there are multiple assembly areas called "assembly islands" on the assembly production line. In the assembly island, the assembly components of the assembly products are the center, and a large number of assembly tools, personnel and materials are arranged around them. Each assembly island has a certain number of parallel teams to handle assembly tasks. In addition, each assembly island handles the work of p skill types in the assembly tasks. The assembly island concentrates the key task nodes and confluence nodes of the process regulations. The critical path of the process flowing through the assembly island can be abstracted as a hybrid flow shop problem model. The reasonable allocation of human resources on the assembly island with parallel teams can be defined as the hybrid flow shop assembly line personnel allocation problem, which can be regarded as a deep coupling problem of the hybrid flow shop scheduling problem (HFSP) and ALWAP.

[0248] In this step, an integer programming model is established based on the product assembly information. The integer programming model includes a first objective function and a second objective function. Among them, the first objective function represents the optimization goal of minimizing the maximum completion time, which measures the economic cost and production efficiency of the enterprise. The second objective function represents the optimization goal of minimizing the unbalanced busy hours, which measures the balance of the workload of the team in the assembly line. Therefore, the establishment of two goals ensures the optimization of the production process from the two aspects of economic benefits and human factors. The integer programming model also includes constraints for instructing assembly workers in the team to perform assembly work.

[0249] Step 102: According to the product assembly information and the integer programming model, a worker allocation code and a task sequence code are determined using a double-layer mixed coding method; the worker allocation code is used to indicate the work position that the assembly worker can go to; the task sequence code is used to indicate the order in which the assembly tasks enter the assembly island.

[0250] In this step, a two-layer hybrid encoding method is proposed to better handle the WAPHAL problem of multivariable coupling. Each chromosome consists of two vectors, namely the worker allocation code and the task order code. The length of the worker allocation code is |W|, where the wth gene bit means the position number assigned to worker w, and the length of the task order code is |J|, and the gene bit order on the code represents the order in which assembly task j enters the critical path of the assembly island.

[0251] Step 103: Based on the integer programming model, multiple product assembly scheduling solutions are obtained according to the worker allocation code and the task sequence code.

[0252] In this step, based on the integer programming model, the above chromosomes are decoded to obtain multiple product assembly scheduling plans. Each assembly scheduling plan includes the order in which assembly tasks are carried out on the assembly island, the team responsible for handling the assembly tasks, and the position in the team to which the assembly worker goes.

[0253] Step 104: using the improved multi-objective improved algorithm MOEA, determine the product assembly scheduling optimization solution from the multiple product assembly scheduling solutions.

[0254] In this step, improvements are made based on the Multi-objective Evolutionary Algorithm (MOEA) framework to obtain an improved multi-objective improvement algorithm MOEA, and the optimal solution is found among multiple product scheduling solutions to obtain the product assembly scheduling optimization solution.

[0255] Optionally, the product assembly information includes:

[0256] Assembly island area information;

[0257] Team information in the assembly island;

[0258] All the skills information required for assembly tasks in the assembly island;

[0259] Assembly worker information;

[0260] Assembly task information;

[0261] The coordinate information of the work positions in the team to which the assembly workers can go;

[0262] the skills that assemblers possess and the proficiency of the skills they possess;

[0263] In an assembly task, the basic working hours required for an assembly worker to complete the assembly task corresponding to his or her skills.

[0264] First, the symbols used in this embodiment and their corresponding meanings are explained:

[0265] I: represents the assembly area set;

[0266] i: represents the index of assembly island area, i∈I;

[0267] k: represents the index of all teams in I, k∈K.

[0268] K: represents the set of all teams in I.

[0269] K i : represents the set of team indexes in assembly island i, U i∈I K i =K.

[0270] p: represents the index of all skills required for the assembly task in I, p∈P.

[0271] P: represents the total skill set required for the assembly task in I.

[0272] P i : represents the skill index set required for the assembly task in assembly island i, U i∈I P i =P.

[0273] w: represents the assembly worker index, w∈W.

[0274] W: represents the set of assembly workers.

[0275] j: represents the assembly task index, j∈J.

[0276] J: represents the assembly task set.

[0277] g w : represents the index of the coordinates of the job position that the assembly worker w can go to, g w ∈G w .

[0278] G w : represents the set of work station coordinates to which assembly worker w can go;

[0279] r wp : represents the proficiency of skill p possessed by assembly worker w;

[0280] f(r wp ): represents the influence of the proficiency of skill p possessed by assembly worker w on the working hours;

[0281] In assembly task j, the quantity is 1, and r wp =1 is the working time required for an assembly worker to complete the assembly task corresponding to skill p, that is, the basic working time.

[0282] Specifically, the assembly island area information is the assembly area set I; the team information in the assembly island includes the set of all teams in all assembly islands I and the set of team indexes in each assembly island i; the information on all skills required for the assembly task in the assembly island includes the set of all skills required for the assembly task in all assembly islands I and the set of skill indexes required for the assembly task in each assembly island i; the assembly worker information is the assembly worker set; the assembly task information is the assembly task set; the job position coordinate information in the team that the assembly worker can go to is the set of job position coordinates that each assembly worker w can go to; the skills and proficiency of the skills possessed by the assembly worker include the skills p possessed by each assembly worker and the proficiency of the skill p possessed by the assembly worker w; in the assembly task, the basic working hours required for the assembly worker to complete the assembly task corresponding to the skill is in the assembly task j, the number is 1, and r wp =1 is the working time required for an assembly worker to complete the assembly task corresponding to skill p, that is, the basic working time.

[0283] The critical process path is planned as four assembly islands i (i∈I) areas. Each assembly task j in the issued assembly task set J must flow through the assembly island in a fixed order. Each assembly task j has different process content and work content related to assembly skill p on the assembly island. The p index set on the assembly island is P i ; Each assembly island has a parallel team to improve assembly efficiency. The k-index set on the assembly island is K i At the same time, corresponding skill positions are set up in the team for assembly skill p to allocate assembly workers. At the team worker position, there are two types of workers: ordinary workers and multi-skilled workers. The total number of ordinary workers is The total number of multi-skilled workers is |W|. In order to highlight the research focus of multi-skilled workers, the embodiment of the present invention will The values ​​are all set to 1; multi-skilled workers have multiple skills and different skill levels. represents the number of ordinary assembly workers with only skill p and proficiency 1 who are pre-placed in team k.

[0284] In assembly task j, a single r wp = 1, the working hours spent by the assembly worker to complete the assembly task of skill type p is set as the basic working hours The assembly task of skill type p determines the length of the team's working time. When processing the work of skill type p of assembly task j, the assembly workers with skill p in team k perform the main assembly work, and the assembly workers without skill p in team k perform auxiliary assembly work. Therefore, the working time t spent on completing assembly task j is reduced. jk is regarded as the total working time of the team members, that is, t jk It represents the working time required for team k to complete the corresponding assembly task in assembly task j. Working time t jk It is the variable working hours, which is mainly related to the task-type influencing factor O and the team-type influencing factor Q. Task-type influencing factor O = (o1, o2, o3, o4, o5, o6, o7, o8, o9, o10, o11, o12), team-type influencing factor Q = (q1, q2, q3), the meanings are as follows:

[0285] o1: Number of fasteners 1; o2: Mass of components; o3: Number of components; o4: Number of structural plates; o5: Number of fasteners 2; o6: Number of electronic components; o7: Number of wrapping layers; o8: Length of pipes; o9: Number of cables; o10: Number of plug-ins 1; o11: Number of plug-ins 2; o12: Length of wrapping; q1: Number of ordinary workers; q2: Number of multi-skilled workers; q3: Skill level of workers.

[0286] In order to facilitate the measurement of time, Δt is set as the discrete unit of working time.

[0287] It should be noted that, in the embodiment of the present invention, the assembly workers are assigned to appropriate teams and a reasonable order for the teams to handle assembly tasks is arranged, given the skills and proficiency of the assembly workers, the assembly tasks and their basic working hours, the skill distribution in the assembly island and the team distribution.

[0288] Optionally, the second objective function is determined based on the mean square error of the maximum working time period of each team and the mean square error of the total working time period between teams.

[0289] Specifically, the first objective function is:

[0290]

[0291] Among them, C k represents the completion time of team k.

[0292] The second objective function is:

[0293] min f 2 =U=U 1 +U 2

[0294] Among them, the minimum busy-idle imbalance U consists of two parts U 1 and U 2 U1 is the mean square error of the maximum working time of each team, which indicates the balance of the longest time for each team to continuously process tasks. 2 is the mean square error of the total working time between teams, which explains the balance of the total workload of each team.

[0295]

[0296]

[0297] If assembly task j is processed by team k, then β jk The value is 1, otherwise β jk is 0; Indicates the end time of the corresponding assembly task processed by team k in assembly task j; Indicates the start time of the corresponding assembly task processed by team k in assembly task j.

[0298] Optionally, the constraint conditions include:

[0299] The first constraint is used to indicate the skills that the assembly workers have and the proficiency of the skills they have;

[0300] The second constraint is used to indicate whether the assembly worker works in a team;

[0301] The third constraint is used to indicate whether the assembly task is processed on the team;

[0302] The fourth constraint is used to indicate whether the team needs to handle the work tasks corresponding to the skill;

[0303] The fifth constraint is used to indicate the processing order of assembly tasks on the team;

[0304] The sixth constraint is used to indicate the degree to which the proficiency of the assembly workers' skills affects the working hours;

[0305] The seventh constraint is used to indicate the working time of the team in processing the assembly task;

[0306] The eighth constraint condition is used to indicate that the assembly worker can be assigned to a team that requires the target skill to handle the corresponding work task only if the proficiency of the target skill possessed by the assembly worker is not 0;

[0307] The ninth constraint is used to indicate the number of assembly workers assigned to all teams;

[0308] The tenth constraint is used to indicate that the same team on the same assembly island must have the same skills;

[0309] The eleventh constraint is used to indicate that the teams on the same assembly island must have the same skills;

[0310] The twelfth constraint is used to indicate that each assembly task can only be assigned to one team on each assembly island;

[0311] The thirteenth constraint is used to indicate that the start time of the assembly task on the team is 0, or the end time of the previous assembly task processed by the team;

[0312] The fourteenth constraint condition is used to indicate that the end time of the assembly task on the team is the sum of the start time of the assembly task on the team and the processing time of the assembly task on the team;

[0313] The fifteenth constraint is used to indicate that an assembly task cannot be processed on multiple assembly islands at the same time;

[0314] The sixteenth constraint is used to indicate that a team can only handle one assembly task at the same time.

[0315] Specifically, the constraints in the integer programming model include:

[0316] The first constraint is:

[0317]

[0318] The above constraints represent the skills that the assembly workers possess and the proficiency of their skills, that is, the level of their skills.

[0319] The second to fifth constraints represent four binary variables.

[0320] Second constraint:

[0321]

[0322] If assembly worker w works in team k, then α wk The value is 1, otherwise α wkis 0.

[0323] The third constraint is:

[0324]

[0325] Among them, on team k, if the processing order of assembly task j is in assembly task j * After , θ jj*k The value is 1, otherwise θ jj*k is 0.

[0326] The fourth constraint is:

[0327]

[0328] If team k needs to handle the assembly task corresponding to skill p, then v kp The value is 1, otherwise v kp The value is 0.

[0329] The fifth constraint:

[0330]

[0331] Among them, in team k, if the processing order of assembly task j is after assembly task j*, then θ jj*k The value is 1, otherwise θ jj*k The value is 0.

[0332] The sixth constraint is:

[0333]

[0334] The above constraints represent the degree to which the proficiency of skill p possessed by assembly worker w affects the working hours.

[0335] The seventh constraint:

[0336]

[0337] The above constraints are used to calculate the working time of team k when processing assembly task j.

[0338] The eighth to twelfth constraints are allocation constraints.

[0339] The eighth constraint is:

[0340]

[0341] The above constraints mean that the assembly worker can be assigned to the team k with skill p only if his proficiency of skill p is not 0.

[0342] Ninth constraint:

[0343]

[0344] The above constraints mean that the sum of the number of assembly workers assigned to all teams is |W|.

[0345] Tenth constraint:

[0346]

[0347] The above constraints mean that the team k on the same assembly island must have the same skills.

[0348] Eleventh constraint:

[0349]

[0350] The above constraints mean that the teams on the same assembly island must have the same skill p.

[0351] The twelfth constraint is:

[0352]

[0353] The above constraints mean that the processing of each assembly task on each assembly island can only be assigned to one team.

[0354] The thirteenth constraint condition and the fourteenth constraint condition are time constraints.

[0355] Thirteenth constraint:

[0356]

[0357] The above constraints indicate that in the initial stage, the start time of the team task should be 0 or the end time of the team's previous task.

[0358] Fourteenth constraint:

[0359]

[0360] The above constraints indicate that the end time of a team task is the sum of the team task start time and the team task processing time.

[0361] The fifteenth constraint condition to the sixteenth constraint condition are task order constraints.

[0362] Fifteenth constraint:

[0363]

[0364] The above constraints mean that a task cannot be processed on multiple assembly islands at the same time (constraints on different stages of the same task).

[0365] The sixteenth constraint is:

[0366]

[0367] The above conditions mean that a team can only handle one task at a time (constraints on different tasks for the same team).

[0368] Optionally, a double-layer hybrid coding method is used to determine the worker allocation code and the task sequence code, including:

[0369] Determine the worker allocation code by integer coding;

[0370] The task sequence code is determined by using an arrangement coding method.

[0371] It should be noted that the WAPHAL problem consists of three sub-problems: the problem of personnel allocation to the team, the problem of arranging the processing order of assembly tasks, and the problem of selecting parallel teams when processing assembly tasks.

[0372] In order to solve the three sub-problems in the scheduling process, a two-layer hybrid coding method is designed. Each chromosome consists of two vectors, namely the worker allocation code and the task order code. The worker allocation code is encoded by integer coding. The task order code is encoded by permutation coding. The schematic diagram of the two-layer hybrid coding is shown in Figure 2 shown.

[0373] The worker allocation code vector is WS, and the worker allocation code length is |W|, where the wth gene bit means the position number to which worker w is assigned, and the range of values ​​on the gene bit means the set of positions G that the worker can go to if constraint 1 is satisfied. w The task sequence code vector is JS, the length of the task sequence code is |J|, and the gene sequence on the code represents the order in which task j enters the critical path of the assembly island.

[0374] Optionally, according to the product assembly information and the integer programming model, a worker allocation code is determined using a double-layer hybrid coding method, including:

[0375] Obtaining required position information according to all skill information required for the assembly task in the assembly island in the product assembly information and the team information in the assembly island in the product assembly information;

[0376] The worker allocation code is determined according to the required job information and the skills of the assembly workers while satisfying the constraint conditions.

[0377] Combine the following Figure 3 , details about worker allocation codes.

[0378] Take the worker allocation code length |W| = 5, and the distribution of skill sets on the assembly island is P 1 =[1,2,3,4],P 2 =[5,6,7], P 3 =[8,9], P 4 =[10,11,12,13], the distribution of the team set on the assembly island is K 1 =[1,2], K 2 =[3,4,5], K 3 =[6,7,8], K 4 =[9,10]. Skills and teams are combined in order to form corresponding positions. Each assembly worker has multiple skills, which determines the positions they can go to. Figure 3 As shown, the second gene value in the WS vector is 2, indicating that worker w=2 is assigned to the available job set G 2 The second position in is (1,2).

[0379] Optionally, based on the integer programming model, multiple product assembly scheduling solutions are obtained according to the worker allocation code and the task sequence code, including:

[0380] Determining first worker information according to the worker allocation code; the first worker information includes allocation information of assembly workers working on the team;

[0381] Determine first working time information according to the first worker information; the first working time information includes the working time of the assembly workers in each team when processing the assembly task;

[0382] Determine task sequence information according to the task sequence code and the first working time information; the task sequence information includes the order in which each assembly task enters each assembly island;

[0383] Determine the first team information according to the first working hour information and the team work information; the first team information is used to indicate that the target assembly task is processed on the target team; the target assembly task is one of each of the assembly tasks; the target team is one of the teams; the team work information includes: the start time of the team processing the current assembly task, the end time of the team processing the current assembly task and the working time of the team processing the assembly task;

[0384] The plurality of product assembly scheduling plans are obtained according to the task sequence information, the first team information and the first worker information.

[0385] The following is a detailed description of chromosome decoding according to the following steps:

[0386] First, the start processing time of all tasks on all shifts and end processing time Set the initial value to 0.

[0387] According to the value of each gene bit of the worker allocation code WS and the set of positions G to which the assembly worker w can go w Determine the position where the assembly worker goes, and thus determine α wk The value of , that is, determine the first worker information. After knowing the worker allocation, calculate the man-hours required for each team to process the assembly task, that is, determine the first man-hour information. Set the task sequence code JS to the initial processing order of the task when it enters the assembly island i=1. After that, enter the subsequent assembly island according to the end processing time of the task, that is, the task sequence information. Calculate the start time of the current task for all teams of each assembly task j Completion time Working time of the team L k And select the team k responsible for processing task j, that is, determine the team work information. According to the first working time information and the team work information, determine the target team for processing the target assembly task, that is, the first team information. The product assembly scheduling plan includes task sequence information, first team information and first worker information.

[0388] Optionally, the method further comprises:

[0389] Determine a first function value of the first objective function corresponding to each of the product assembly scheduling plans;

[0390] Determine a second function value of the second objective function corresponding to each of the product assembly scheduling plans.

[0391] It should be noted that when decoding, the multi-objective function value f of the individual (product assembly scheduling plan) is calculated. 1 (the first function value of the first objective function), f 2 (The second function value of the second objective function). And output the start time and completion time set of all assembly tasks in each team.

[0392] Optionally, determining first team information according to the first working hour information and team work information includes:

[0393] Determine, among each of the teams, the first team that has the closest end time for processing the current assembly task as the target team;

[0394] In the case where there are multiple first shifts, among the first shifts, the second shift with the shortest working time is determined as the target shift;

[0395] In the case that there are multiple second teams, the third team in the second teams is determined as the target team; the third team is any one of the second teams.

[0396] According to the first working hour information and the team work information, the process of determining the target team for processing the target assembly task is as follows:

[0397] Select the target team to handle the target assembly task according to the Idle-Rule. That is, the current completion time is given priority. The smallest team; if the completion time If they are the same, select the working time L k The smallest team; if the working time is L k If the values ​​of the two groups are the same, the team is randomly selected.

[0398] Optionally, according to the improved multi-objective improved algorithm MOEA, before the multiple product assembly scheduling schemes determine the product assembly scheduling optimization scheme, the method further includes:

[0399] Through the improvement strategy, the multi-objective improved algorithm MOEA is improved to obtain the improved multi-objective improved algorithm MOEA;

[0400] The improvement strategy includes at least one of the following:

[0401] Improved search mechanism strategies;

[0402] Improved strategies for evolving population structure;

[0403] Improve the population distribution maintenance mechanism strategy.

[0404] It should be noted that since Schaffer proposed the Vector Evaluated Genetic Algorithm (VEGA) in 1984, researchers have continuously proposed new multi-objective evolutionary algorithms (MOEA), which can be roughly divided into three categories: MOEA based on decomposition strategy, MOEA based on dominance relationship, and MOEA based on evaluation index. Among them, MOEA based on dominance relationship has the advantages of fast calculation speed and easy implementation. Therefore, improvements of three different strategies are realized under the framework of MOEA algorithm based on dominance relationship. The overall framework includes encoding, population initialization, population selection, individual crossover, individual mutation, fitness evaluation, non-dominated solution set construction, offspring distribution maintenance, stopping criteria, and decoding.

[0405] The embodiment of the present invention, based on the MOEA algorithm framework, proposes three improvement strategies in the three aspects of search mechanism, evolutionary population structure, and population distribution maintenance mechanism.

[0406] Before introducing the three improvement strategies in detail, the following are explained:

[0407] Since the worker allocation code is set to real integer encoding and the task order code is set to permutation encoding, it is necessary to set the crossover mutation methods that are suitable for the encoding methods. The worker allocation code crossover method is set to uniform distribution crossover Xovud, and the mutation method is set to two-point swap mutation Mutswap. The task order code crossover method is set to sequential crossover Xovox, and the mutation method is set to permutation encoding Mutpp.

[0408] The embodiment of the present invention improves the strategy of maintaining offspring distribution in a multi-objective evolutionary algorithm, and common strategies include crowding distance, clustering, adaptive grid, etc. Three different improvement strategies are provided for selection under different working conditions.

[0409] Optionally, the multi-objective improved algorithm MOEA is improved by improving the search mechanism strategy to obtain an improved multi-objective improved algorithm MOEA, including:

[0410] According to a preset mutation mode suitable for permutation coding and a preset mutation mode suitable for integer coding, a mutation combination complex scheme is obtained;

[0411] According to the mutation combination scheme, the evolved offspring are further searched to obtain the mutant offspring, so that the improved multi-objective improved algorithm MOEA searches among the evolved offspring, the mutant offspring and the population parent generation.

[0412] Improved strategy 1, that is, using the mutation pool to evolve offspring O in the improved search mechanism strategy g Further search is performed to generate mutant progeny M g , in the mutant progeny M g , Evolutionary O g and the parent population C g The mutation pool is a set of mutation methods for chromosomes, which is composed of a set of mutation methods for permutation coding and a set of mutation methods for integer coding. PM = {fragment reversal, fragment shift, permutation coding, two-point swap}, a set of mutation methods suitable for integer coding S IM = {BGA algorithm mutation, differential mutation, Gaussian mutation, segment reversal mutation, segment shift mutation, polynomial mutation, two-point swap mutation, uniform mutation}, a total of 8 combination schemes S IM ×S PM ={(S IM [0],S PM [0]),(S IM [1],S PM[1]),(S IM [2],S PM [2]),(S IM [3],S PM [3]),(S IM [0],S PM [4]),(S IM [1],S PM [5]),(S IM [2],S PM [6]),(S IM [3],S PM [7])}. In line (4) of the pseudocode of Algorithm 3.1, we traverse the combination S IM ×S PM The mutation scheme S in the evolutionary offspring O is g Mutation A total of 8 mutant offspring were generated

[0413] Optionally, the multi-objective improved algorithm MOEA is improved by improving the evolutionary population structure strategy to obtain an improved multi-objective improved algorithm MOEA, including:

[0414] Divide the evolutionary offspring into non-dominant hierarchies;

[0415] Determine the evolutionary offspring whose non-dominant level is less than or equal to the first preset level as the elite offspring;

[0416] The evolutionary offspring are searched according to the elite offspring to obtain the mutant elite offspring, so that the improved multi-objective improved algorithm MOEA searches in the mutant elite offspring.

[0417] Improved strategy 2, that is, in the improved evolutionary population structure strategy, the ENS_SS strategy is first used to divide the evolutionary offspring into non-dominated levels, and the elite offspring E whose non-dominated level is less than or equal to EL (the first preset level) among the evolutionary offspring is selected. g Put into the mutation pool for mutation search to generate 8 mutant elite offspring Searching for elite solution sets can greatly improve the efficiency of the algorithm.

[0418] Optionally, the multi-objective improved algorithm MOEA is improved by improving the population distribution maintaining mechanism strategy to obtain an improved multi-objective improved algorithm MOEA, including:

[0419] Divide the evolutionary offspring into non-dominant hierarchies;

[0420] According to the non-kinship calculation function, the crowding distance strategy is improved so that the improved multi-objective improved algorithm MOEA generates the next generation of individuals according to the evolutionary offspring of the target non-dominated hierarchy;

[0421] Wherein, the non-kinship calculation function is determined based on the crowding distance and the similarity;

[0422] The target non-dominated hierarchy is a non-dominated hierarchy in which the total number of individuals is less than the preset number of individuals in the population, and the total number of individuals after merging is greater than the preset number of individuals in the population.

[0423] Improved strategy 3, that is, the improved population distribution maintenance mechanism strategy improves the crowding distance strategy in NSGA-Ⅱ and proposes a non-close relative degree calculation formula. Specifically, the following formula is the crowding distance formula for the two objectives:

[0424] I n .d1=|I n+1 .f 1 -I n-1 .f 1 |+|I n+1 .f 2 -I n-1 .f 2 |

[0425] Where d1 represents the crowding distance between two targets.

[0426] The following is the similarity calculation formula:

[0427]

[0428] Among them, d2 represents the similarity between two targets.

[0429] The similarity calculation formula is calculated for each gene position m of the chromosome to evaluate the distance between gene positions m of different chromosomes.

[0430] The following is the formula for calculating non-kinship:

[0431] I n .D=I n .d1+I n .d2

[0432] Where D represents the degree of non-kinship.

[0433] Non-kinship is the sum of crowding distance and similarity, which can enhance the diversity of offspring populations in terms of fitness and phenotype. n .f′ m Meaning: Individual I n f′ m Gene value.

[0434] The main calculation steps of the improved multi-objective improved algorithm MOEA are as follows:

[0435] Randomly generate an initial population C with Nind as the number of individuals 0 , C 0 Enter the evolutionary iterative process to become C g . For the population C in iteration g Perform evolution operations including crossover and mutation to generate the evolved subpopulation O g If improved strategy 1 and improved strategy 2 are used, the population O g Generate a set of mutant offspring M g On the contrary, M g is an empty set. For the population C before evolution g , the evolved subpopulation O g and the mutant offspring set M g Perform decoding operations and calculate their multi-objective function values. According to the multi-objective function values, use the ENS_SS strategy to decode C g , O g and M g All individuals in the non-dominated order are sorted and divided into multiple layers according to the non-dominated level. Select the non-dominated layer set F whose total number of individuals is just less than the number of individuals in the population Nind m , and F m After the merger, the total number of individuals will be greater than Nind's critical layer F c . For the critical layer F c Consider whether to use improved strategy 3. If used, generate based on strategy 3 If not used, it is generated based on crowding distance sorting. Gradually remove The last individual in F m and The total number of individuals in is Nind. At this time, the next generation population C g+1 F m and The loop is iterated until the stopping criterion is met, that is, the number of iterations reaches the maximum number of iterations MAXGEN, and the final solution C is output. MAXGEN .

[0436] Optionally, the method further comprises:

[0437] Analyze the complexity of the improved first multi-objective improved algorithm MOEA to obtain a first complexity analysis result; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved search mechanism strategy;

[0438] Analyze the complexity of the improved second multi-objective improved algorithm MOEA to obtain a second complexity analysis result; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved evolutionary population structure strategy;

[0439] Analyze the complexity of the improved third multi-objective improved algorithm MOEA to obtain a third complexity analysis result; the improved third multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved population distribution maintaining mechanism strategy;

[0440] Analyze the complexity of the preset multi-objective improved algorithm to obtain a fourth complexity analysis result;

[0441] The first complexity analysis result, the second complexity analysis result, the third complexity analysis result and the fourth complexity analysis result are compared to obtain a first comparison result indicating the complexity of the improved multi-objective improved algorithm MOEA.

[0442] The following is an analysis of the algorithm complexity of the multi-objective improved algorithm MOEA and the classic NSGA2 algorithm in the MOEA algorithm improved by three improvement strategies:

[0443] The main time cost of NSGA2 (preset multi-objective improved algorithm) with r dimensions and population size N is in the construction of the non-dominated layer, and its total time complexity is O(r(2N) 2 ). See Figure 4 In the improved strategy 1, the evolutionary offspring O g 8 mutations were performed, resulting in 8 mutant offspring Therefore, the object of non-dominated sorting after merging is a population of 10N individuals, and its complexity is O(r(10N) 2 ), which is 25 times faster than the NSGA2 algorithm. Figure 5 In improved strategy 2, firstly, the evolutionary offspring O g A non-dominated sort is performed, that is, the time complexity is O(rN 2 ), the elite individuals are selected. Generally, EL is set to ensure that the number of elite individuals is less than 0.2N. Therefore, the object of non-dominated sorting after merging is a population with a number of individuals less than 2N+0.2*8N=3.6N. Therefore, the maximum total time complexity is about O(r14N 2 ), which is 3.5 times faster than the NSGA2 algorithm. Figure 6In improved strategy 3, the algorithm complexity comes from the calculation of non-kinship. The complexity of calculating the crowding distance in the NSGA2 algorithm is O(r(2N)log(2N))(26), and the complexity of calculating the non-kinship is O((r+|W|+|J|)*(2N)*log(2N)). Therefore, in terms of the order of magnitude of algorithm complexity, the total time complexity of the three improved strategies and NSGA2 is O(rN 2 ).

[0444] Optionally, the method further comprises:

[0445] Obtaining a first assembly scheduling test plan according to a preset number of first assembly islands, a preset number of first assembly tasks, and a preset number of first assembly workers;

[0446] By using the improved first multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, a first union of non-dominated solutions and a first Pareto front of non-dominated solutions in the first union are obtained, wherein the first Pareto front is used to indicate the ability of the improved first multi-objective optimization algorithm MOEA to approach the optimal solution; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved search mechanism strategy;

[0447] By using the improved second multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, a second union of non-dominated solutions and a second Pareto front of non-dominated solutions in the second union are obtained, wherein the second Pareto front is used to indicate the ability of the improved second multi-objective optimization algorithm MOEA to approach the optimal solution; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0448] By using the improved fourth multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, a third union of non-dominated solutions and a third Pareto front of non-dominated solutions in the third union are obtained, wherein the third Pareto front is used to indicate the ability of the improved fourth multi-objective optimization algorithm MOEA to approach the optimal solution; the improved fourth multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0449] Obtaining, by using a preset multi-objective optimization algorithm and the first assembly scheduling test scheme, a fourth union of non-dominated solutions and a fourth Pareto front of the non-dominated solutions in the fourth union, wherein the fourth Pareto front is used to indicate the ability of the preset multi-objective optimization algorithm to approach an optimal solution;

[0450] The first Pareto front, the second Pareto front, the third Pareto front and the fourth Pareto front are compared to obtain a second comparison result indicating the ability of the improved multi-objective optimization algorithm MOEA to approach the optimal solution.

[0451] In this embodiment of the present invention, basic working hours are generated according to production examples. The test instance is , and = 10min is used as a discrete time unit. After declassification, the main work details of the critical path assembly process of aerospace products are classified into 13 basic ∈ skills, and the distribution index of each skill in the assembly island is P1 = [1,2,3,4], P2 = [5,6,7], P3 = [8,9], P4 = [10,11,12,13], such as Figure 7 shown.

[0452] The skill proficiency of multi-skilled workers is set to a random number in the interval [0,3] to generate personnel test instance data (the first assembly scheduling test plan). A skill proficiency of 0 means that the assembler does not have this skill. The number of assembly islands |I| = 4. The number of teams |K| = 10. There are 5 distribution schemes in the number of assembly islands, namely {L1 = [2,3,2,3], L2 = [3,2,2,3], L3 = [2,2,3,3], L4 = [3,3,2,2], L5 = [3,2,2,3]}. The number of assembly tasks is set to two scales, namely {|J1| = 30, |J2| = 45}, and the number of multi-skilled assemblers is set to two scales, namely {|W1| = 45, |W2| = 60}. In this way, 20 groups of test instances of different scales and team distributions are generated. For example, I-321 is represented as a test instance composed of L4, |J2|=45 and |W1|=45.

[0453] In order to compare the performance of the three improved strategies and the six preset algorithms (preset multi-objective optimization algorithms) horizontally, the parameters of all algorithms are set as shown in Table 1. The number of assembly islands is 4. Each individual population size (Nind) is set to 100 individuals, and the maximum number of algorithm iterations MAXGEN is set to 100 times. In the hybrid coding, the same crossover probability of Worker Allocation code and Task Order code are used, and the crossover probability is set to 0.8, and the mutation probability is set to the inverse of the number of gene bits, that is, 1 / |W| and 1 / |J|. The number of mutant offspring in the improved strategy is set to 8, the number of elite screening levels in the improved strategy 2 is set to 6, and the scaling factor parameter F of the differential mutation operator in the mutation pool is set to 0.5.

[0454] Table 1 Parameters of optimization algorithm

[0455] Nind MAXGEN |I| PC_WA PM_WA PC_TO PM_TO NMO EL F 100 100 4 0.8 1 / |W| 0.8 1 / |J| 8 6 0.5

[0456] The three improved strategies are represented by EM1, EM2, and EM3 respectively, and the six preset algorithms include: NSGA2 algorithm, awGA algorithm, NSGA2_archive algorithm, NSGA3 algorithm, RVEA algorithm, and RVEA_RES algorithm.

[0457] The mutation evolution algorithms EM1, EM2, and EM3 generated by three improved strategies were compared with NSGA2, awGA, NSGA2_archive, NSGA3, RVEA, and RVEA_RES algorithms. Among them, EM1 used improved strategy 1, EM2 used improved strategy 2, and EM3 used improved strategies 2 and 3. awGA is an adaptive weight method multi-objective aggregation function that solves multi-objective problems by aggregating multiple objectives into a single objective with varying weights. RVEA is a reference vector-based evolutionary algorithm proposed for high-dimensional problems. Compared with the RVEA algorithm, the RVEA_RES algorithm adds a regeneration strategy with reference points. NSGA2_archive is an NSGA algorithm with global archive. For ease of representation, the algorithms are numbered from A0 to A8 in the following table. In order to ensure the generalization and reduce randomness of the algorithm comparison results, each algorithm was run 30 times, and the experimental data were processed by different test methods.

[0458] Table 2 Algorithm and number comparison table

[0459]

[0460] We now specifically analyze the ability of these nine algorithms to approach the optimal solution in the scatter plot of 20 examples. The specific operation is to run the algorithm 30 times, then take the union of the resulting non-dominated solution sets, and extract the non-dominated solution frontier PF (Pareto front) in the union to represent the ability of the algorithm to approach the optimal solution. Figures 8 to 11 From the PF scatter plots comparing the nine algorithms generated under the test instances I-111, I-112, I-121, and I-122, it can be seen that as the first objective function value f1 obtained by algorithms EM1, EM2, and EM3 increases, the second objective function value f2 remains at a relatively small value, indicating that the ability of algorithms EM1, EM2, and EM3 to approximate the optimal solution is better than that of the other six preset algorithms.

[0461] Optionally, the method further comprises:

[0462] Obtaining a second assembly scheduling test plan according to a preset number of second assembly islands, a preset number of second assembly tasks, and a preset number of second assembly workers;

[0463] By using the improved first multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, a first union of non-dominated solutions is obtained, and according to the first union, a first inverse generation distance analysis result and a first hypervolume index analysis result of the improved first multi-objective optimization algorithm MOEA are obtained; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved search mechanism strategy;

[0464] By using the improved second multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, a second union of non-dominated solutions is obtained, and according to the second union, a second inverse generation distance analysis result and a second hypervolume index analysis result of the improved second multi-objective optimization algorithm MOEA are obtained; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0465] By using the improved fourth multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, a third union of non-dominated solutions is obtained, and according to the third union, a third inverse generation distance analysis result and a third hypervolume index analysis result of the improved fourth multi-objective optimization algorithm MOEA are obtained; the improved fourth multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0466] Obtaining a fourth union of non-dominated solutions through a preset multi-objective optimization algorithm and the second assembly scheduling test scheme, and obtaining a fourth inverse generation distance analysis result and a fourth hypervolume index analysis result of the preset multi-objective optimization algorithm according to the third union;

[0467] Comparing the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result and the fourth inverse generation distance analysis result to obtain a third comparison result indicating the convergence of the improved multi-objective optimization algorithm MOEA;

[0468] The first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result are compared to obtain a fourth comparison result indicating the convergence of the improved multi-objective optimization algorithm MOEA.

[0469] The embodiment of the present invention further compares the performance of three improved strategies and six preset algorithms (preset multi-objective optimization algorithms) horizontally.

[0470] The three improved strategies are represented by EM1, EM2, and EM3 respectively, and the six preset algorithms include: NSGA2 algorithm, awGA algorithm, NSGA2_archive algorithm, NSGA3 algorithm, RVEA algorithm, and RVEA_RES algorithm.

[0471] The mutation evolution algorithms EM1, EM2, and EM3 generated by three improved strategies were compared with NSGA2, awGA, NSGA2_archive, NSGA3, RVEA, and RVEA_RES algorithms. EM1 used improved strategy 1, EM2 used improved strategy 2, and EM3 used improved strategies 2 and 3.

[0472] The second assembly scheduling test plan is consistent with the first assembly scheduling test plan and will not be described in detail here.

[0473] The two indicators of inverted generational distance (IGD) and hypervolume (HV) are used to evaluate the distribution breadth, convergence and diversity of solutions. The higher the quality of the solution, the smaller the IGD value and the larger the HV value. The IGD indicator requires the establishment of a PF reference set, and the union of all non-dominated solution sets generated by the 9 algorithms is taken as the PF reference set. The HV indicator requires the establishment of a reference point set, and the non-dominated solution set generated by NSGA2 is used as the reference point set. According to the test instances I-111, I-112, I-121 and I-122, the aggregated values ​​of the two indicators of each generation of the algorithm are obtained. The aggregation method is to calculate the two indicator values ​​of the non-dominated solution set generated in each generation each time the algorithm is run, and then sum and average the 30 IGD values ​​and 30 HV values ​​of each generation to represent the average performance when the algorithm is running.

[0474] Take the IGD value indicator as an example. Figures 12 to 15 , the algorithms basically complete the search for high-quality solutions between 20 and 40 generations, and EM1, EM2 and EM3 converge the fastest. When iterating to the last 80 to 100 generations, the values ​​converged to by the proposed EM1, EM2 and EM3 are significantly smaller than the values ​​converged to by the other 6 preset algorithms. Therefore, the three proposed improvement strategies are significantly better than the other 6 preset algorithms. Among the other algorithms, the RVEA_RES algorithm is better than the other algorithms, and the NSGA3 algorithm performs the worst. Let's take the HV value indicator as an example for analysis. Please refer to Figures 16 to 19 , the algorithms basically complete the search for high-quality solutions between 20 and 60 generations, and EM1, EM2 and EM3 have the fastest divergence speed. When iterating to the last 80 to 100 generations, the values ​​diverged by EM1, EM2 and EM3 are significantly smaller than the values ​​converged by the other 6 preset algorithms. Therefore, the three proposed improvement strategies are significantly better than the other 6 preset algorithms.

[0475] Optionally, the method further comprises:

[0476] Obtaining a fifth comparison result indicating the stability of the improved multi-objective optimization algorithm MOEA according to the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result, and the fourth inverse generation distance analysis result;

[0477] According to the first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result, a sixth comparison result indicating the stability of the improved multi-objective optimization algorithm MOEA is obtained.

[0478] Furthermore, the embodiment of the present invention also calculates the IGD and HV indicators for the final non-dominated solution set generated after the 9 algorithms are independently run 30 times, forming a total of 60 indicator data, as follows Figure 20 to Figure 23 The box plot of the IGD indicator shown and Figure 24 to Figure 27 The box plot of the HV index is shown in the figure. It can be seen that the median IGD of EM1, EM2 and EM3 is smaller than that of other algorithms, and the median HV is larger than that of other algorithms, which indicates that the quality of the solutions of EM1, EM2 and EM3 algorithms is the highest. At the same time, EM1, EM2 and EM3 have a shorter interquartile range than other algorithms, which indicates that EM1, EM2 and EM3 have good algorithm stability.

[0479] Optionally, the method further comprises:

[0480] Taking the improved first multi-objective optimization algorithm MOEA as a benchmark, obtaining a seventh comparison result indicating the dominance of the improved multi-objective optimization algorithm MOEA according to the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result and the fourth inverse generation distance analysis result;

[0481] Taking the improved first multi-objective optimization algorithm MOEA as a benchmark, according to the first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result, an eighth comparison result indicating the dominance of the improved multi-objective optimization algorithm MOEA is obtained.

[0482] The Wilcoxon sign test with a significance level of 0.05 was used to perform statistical analysis on the 60 indicator data, with algorithm A0 (EM1) as the benchmark. The HV indicator data was inversely manipulated, and the data with asterisks indicate that algorithm A0 dominates the algorithm. The results of the 20 test instances are shown in Table 3 below.

[0483] Table 3 Analysis results of test examples

[0484]

[0485] It can be seen from Table 3 that the A0 algorithm completely dominates the A3 to A8 algorithms. The huge difference in order of magnitude between the A1 and A2 algorithms and the A3 to A8 algorithms also shows that the A1 and A2 algorithms completely dominate the A3 to A8 algorithms.

[0486] The embodiment of the present invention specifically analyzes the advantages and disadvantages of the above three improvement strategies in 20 test examples. In the indicator trend chart, there is no obvious distinction between the three algorithms. Now, the data in the box plot, PF scatter plot and Wilcoxon sign test are statistically analyzed.

[0487] In the Wilcoxon sign test, the IGD index of the A0 algorithm on the example I-522 and the HV index on the examples I-211 and I-522 dominate the A1 algorithm, the IGD index on the examples I-112, I-211, I-521 and I-522 and the HV index on the examples I-122, I-211, I-521 and I-522 dominate the A2 algorithm, and the HV index on the example I-511 is dominated by the algorithm A1.

[0488] Optionally, the method further comprises:

[0489] According to the first Pareto front, the second Pareto front and the third Pareto front, a ninth comparison result indicating the applicability of the improved multi-objective optimization algorithm MOEA is obtained.

[0490] exist Figures 8 to 11 In the PF scatter plot shown, in test instance I-111, A1>A0>A2; in test instance I-112, A0≈A1>A2; in test instance I-121, there is no obvious difference between the three improvement strategies; in test instance I-122, A0>A1≈A2. The statistical analysis of 20 groups of test instances in the above manner is shown in Table 4 below, among which the A0 algorithm is the best 13 times, 4 times medium, and 3 times poor; the A1 algorithm is the best 7 times, 7 times medium, and 6 times poor; the A2 algorithm is the best 8 times, 8 times medium, and 4 times poor. Analysis of the total score shows that A0>A2>A1. In Table 4, "1" means the best, "2" means medium, "3" means poor, and "sum" means there is no obvious difference.

[0491] Table 4 Test example statistical analysis Table 1

[0492]

[0493] Optionally, the method further comprises:

[0494] According to the first inverse generation distance analysis result, the second inverse generation distance analysis result and the third inverse generation distance analysis result, a tenth comparison result indicating the applicability of the improved multi-objective optimization algorithm MOEA is obtained.

[0495] The method further comprises:

[0496] According to the first hypervolume index analysis result, the second hypervolume index analysis result and the third hypervolume index analysis result, an eleventh comparison result is obtained for indicating the applicability of the improved multi-objective optimization algorithm MOEA.

[0497] exist Figure 20 to Figure 23 The box plot of the IGD indicator shown and Figure 24 to Figure 27 In the box plot of the HV index shown in the figure, the A0, A1 and A2 algorithms are scored and evaluated from the two aspects of the median and the interquartile range. "1" means the best, "2" means medium, "3" means poor, and "sum" means there is no obvious difference, and the same score is taken for the same. The statistical results are shown in Tables 5 and 6 below. The total scores of the A0, A1 and A2 algorithms are analyzed in Table 5 below. In the IGD index, the median and the interquartile range are both A0>A1>A2; the total scores of the A0, A1 and A2 algorithms are analyzed in Table 6 below. In the HV index, the median is A0>A1>A2, and the interquartile range is A1>A2>A0.

[0498] Table 5 Test example statistical analysis Table 2

[0499]

[0500] Table 6 Test example statistical analysis Table 3

[0501]

[0502] The embodiment of the present invention proposes a new WAPHAL problem for the HFSP and WAP problems, which effectively expands the hybrid flow shop problem and the worker dispatching problem; for the WAPHAL problem, a corresponding integer programming model is established with the maximum completion time and the busy-idle balance of workers as optimization objectives, and a hybrid encoding method of worker allocation code and task order code and a decoding method thereof are designed; based on the MOEA algorithm framework, three improvement strategies are proposed from three aspects: search mechanism, evolutionary population structure, and population distribution maintenance mechanism to improve the quality of the solution.

[0503] The WAPHAL problem in the embodiment of the present invention comes from the production line of multi-variety and small-batch aerospace products, aiming to improve production efficiency and balance the workload of workers. The algorithms EM1, EM2, and EM3 formed based on the three improvement strategies are compared with the NSGA2, awGA, NSGA2_archive, NSGA3, RVEA, and RVEA_RES algorithms. Under 20 test instances, the Pareto solutions generated by the proposed EM1, EM2, and EM3 algorithms dominate the other six MOEA algorithms in terms of the ability to approach the optimal solution, the HV and IGD indicators, and the convergence speed.

[0504] like Fig.28 As shown, an embodiment of the present invention further provides a product assembly scheduling device, comprising:

[0505] Model building module 2801, used to build an integer programming model according to product assembly information; the integer programming model includes a first objective function for indicating the maximum value of minimizing the completion time of the team, a second objective function for indicating minimizing the busy-idle imbalance, and constraints for indicating the assembly workers in the team to perform assembly work;

[0506] The first determination module 2802 is used to determine the worker allocation code and the task sequence code according to the product assembly information and the integer programming model by using a double-layer mixed coding method; the worker allocation code is used to indicate the work position that the assembly worker can go to; the task sequence code is used to indicate the order in which the assembly tasks enter the assembly island;

[0507] A second determination module 2803 is used to obtain multiple product assembly scheduling solutions based on the integer programming model, according to the worker allocation code and the task sequence code;

[0508] The third determination module 2804 is used to determine the product assembly scheduling optimization solution from the multiple product assembly scheduling solutions by using the improved multi-objective improved algorithm MOEA.

[0509] In an embodiment of the present invention, based on product assembly information, an integer programming model is established, which includes a first objective function for indicating the maximum value of minimizing the team completion time, a second objective function for indicating minimizing the busy-idle imbalance, and an integer programming model for indicating the constraints for assembly workers in the team to perform assembly work. Minimizing the maximum completion time and minimizing the busy-idle imbalance are used as optimization goals, and a plurality of product assembly scheduling schemes and an improved multi-objective improved algorithm MOEA are obtained through hybrid encoding and decoding to obtain a product assembly scheduling optimization scheme, which can improve production efficiency and reduce the occurrence of busy-idle imbalance among assembly workers.

[0510] Optionally, the product assembly information includes:

[0511] Assembly island area information;

[0512] Team information in the assembly island;

[0513] All the skills information required for assembly tasks in the assembly island;

[0514] Assembly worker information;

[0515] Assembly task information;

[0516] The coordinate information of the work positions in the team to which the assembly workers can go;

[0517] the skills that assemblers possess and the proficiency of the skills they possess;

[0518] In an assembly task, the basic working hours required for an assembly worker to complete the assembly task corresponding to his or her skills.

[0519] Optionally, the second objective function is determined based on the mean square error of the maximum working time period of each team and the mean square error of the total working time period between teams.

[0520] Optionally, the constraint conditions include:

[0521] The first constraint is used to indicate the skills that the assembly workers have and the proficiency of the skills they have;

[0522] The second constraint is used to indicate whether the assembly worker works in a team;

[0523] The third constraint is used to indicate whether the assembly task is processed on the team;

[0524] The fourth constraint is used to indicate whether the team needs to handle the work tasks corresponding to the skill;

[0525] The fifth constraint is used to indicate the processing order of assembly tasks on the team;

[0526] The sixth constraint is used to indicate the degree to which the proficiency of the assembly workers' skills affects the working hours;

[0527] The seventh constraint is used to indicate the working time of the team in processing the assembly task;

[0528] The eighth constraint condition is used to indicate that the assembly worker can be assigned to a team that requires the target skill to handle the corresponding work task only if the proficiency of the target skill possessed by the assembly worker is not 0;

[0529] The ninth constraint is used to indicate the number of assembly workers assigned to all teams;

[0530] The tenth constraint is used to indicate that the same team on the same assembly island must have the same skills;

[0531] The eleventh constraint is used to indicate the different skills that different teams on the same assembly island must possess;

[0532] The twelfth constraint is used to indicate that each assembly task can only be assigned to one team on each assembly island;

[0533] The thirteenth constraint is used to indicate that the start time of the assembly task on the team is 0, or the end time of the previous assembly task processed by the team;

[0534] The fourteenth constraint condition is used to indicate that the end time of the assembly task on the team is the sum of the start time of the assembly task on the team and the processing time of the assembly task on the team;

[0535] The fifteenth constraint is used to indicate that an assembly task cannot be processed on multiple assembly islands at the same time;

[0536] The sixteenth constraint is used to indicate that a team can only handle one assembly task at the same time.

[0537] Optionally, the first determining module 2802 includes:

[0538] A first determining unit, configured to determine the worker allocation code by using integer coding;

[0539] The second determining unit is used to determine the task sequence code by using an arrangement coding method.

[0540] Optionally, the first determining module 2802 includes:

[0541] A third determining unit is used to obtain required position information according to all skill information required for the assembly task in the assembly island in the product assembly information and the team information in the assembly island in the product assembly information;

[0542] The fourth determining unit is used to determine the worker allocation code according to the required job information and the skills of the assembly worker while satisfying the constraint condition.

[0543] Optionally, the second determining module 2803 includes:

[0544] a fifth determining unit, configured to determine first worker information according to the worker allocation code; the first worker information includes allocation information of assembly workers working on the team;

[0545] a sixth determining unit, configured to determine first working time information according to the first worker information; the first working time information includes working hours of assembly workers in each team when processing assembly tasks;

[0546] a seventh determining unit, configured to determine task sequence information according to the task sequence code and the first working time information; the task sequence information includes the order in which each assembly task enters each assembly island;

[0547] an eighth determination unit, configured to determine first team information according to the first working hour information and team work information; the first team information is used to indicate that a target assembly task is processed on a target team; the target assembly task is one of each of the assembly tasks; the target team is one of the teams; the team work information includes: a start time when the team processes the current assembly task, an end time when the team processes the current assembly task, and a working time for the team to process the assembly task;

[0548] A ninth determining unit is used to obtain the plurality of product assembly scheduling plans according to the task sequence information, the first team information and the first worker information.

[0549] Optionally, the second determining module 2803 further includes:

[0550] A tenth determining unit, configured to determine a first function value of the first objective function corresponding to each of the product assembly scheduling plans;

[0551] An eleventh determining unit is used to determine a second function value of the second objective function corresponding to each of the product assembly scheduling plans.

[0552] Optionally, the eighth determining unit is specifically configured to:

[0553] Determine, among each of the teams, the first team that has the closest end time for processing the current assembly task as the target team;

[0554] In the case where there are multiple first shifts, among the first shifts, the second shift with the shortest working time is determined as the target shift;

[0555] In the case that there are multiple second teams, the third team in the second teams is determined as the target team; the third team is any one of the second teams.

[0556] Optionally, the device further comprises:

[0557] An algorithm improvement module is used to improve the multi-objective improved algorithm MOEA through an improvement strategy to obtain an improved multi-objective improved algorithm MOEA;

[0558] The improvement strategy includes at least one of the following:

[0559] Improved search mechanism strategies;

[0560] Improved strategies for evolving population structure;

[0561] Improve the population distribution maintenance mechanism strategy.

[0562] Optionally, the algorithm improvement module includes:

[0563] A first processing unit, configured to obtain a mutation combination complex scheme according to a preset mutation mode suitable for permutation coding and a preset mutation mode suitable for integer coding;

[0564] The first search unit is used to further search the evolved offspring according to the mutation combination scheme to obtain the mutant offspring, so that the improved multi-objective improved algorithm MOEA searches among the evolved offspring, the mutant offspring and the population parent generation.

[0565] Optionally, the algorithm improvement module includes:

[0566] The first level division unit is used to divide the evolutionary offspring into non-dominated levels;

[0567] A twelfth determination unit, used to determine the evolutionary offspring whose non-dominant level is less than or equal to the first preset level as the elite offspring;

[0568] The second search unit is used to search the evolutionary offspring according to the elite offspring to obtain the mutant elite offspring, so that the improved multi-objective improved algorithm MOEA searches in the mutant elite offspring.

[0569] Optionally, the algorithm improvement module includes:

[0570] The second level division unit is used to divide the evolutionary offspring into non-dominated levels;

[0571] The second processing unit is used to improve the crowding distance strategy according to the non-close relative degree calculation function, so that the improved multi-objective improved algorithm MOEA generates the next generation of individuals according to the evolutionary offspring of the target non-dominated level;

[0572] Wherein, the non-kinship calculation function is determined based on the crowding distance and the similarity;

[0573] The target non-dominated hierarchy is a non-dominated hierarchy in which the total number of individuals is less than the preset number of individuals in the population, and the total number of individuals after merging is greater than the preset number of individuals in the population.

[0574] Optionally, the algorithm improvement module further includes:

[0575] A first analysis unit is used to analyze the complexity of the improved first multi-objective improved algorithm MOEA to obtain a first complexity analysis result; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved search mechanism strategy;

[0576] The second analysis unit is used to analyze the complexity of the improved second multi-objective improved algorithm MOEA to obtain a second complexity analysis result; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved evolutionary population structure strategy;

[0577] The third analysis unit is used to analyze the complexity of the improved third multi-objective improved algorithm MOEA to obtain a third complexity analysis result; the improved third multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved population distribution maintaining mechanism strategy;

[0578] A fourth analysis unit, used to analyze the complexity of the preset multi-objective improved algorithm to obtain a fourth complexity analysis result;

[0579] The first comparison unit is used to compare the first complexity analysis result, the second complexity analysis result, the third complexity analysis result and the fourth complexity analysis result to obtain a first comparison result indicating the complexity of the improved multi-objective improved algorithm MOEA.

[0580] Optionally, the algorithm improvement module further includes:

[0581] A thirteenth determining unit, configured to obtain a first assembly scheduling test plan according to a preset number of first assembly islands, a preset number of first assembly tasks, and a preset number of first assembly workers;

[0582] A third processing unit is used to obtain a first union of non-dominated solutions and a first Pareto front of non-dominated solutions in the first union through an improved first multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, wherein the first Pareto front is used to indicate the ability of the improved first multi-objective optimization algorithm MOEA to approach an optimal solution; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved search mechanism strategy;

[0583] a fourth processing unit, configured to obtain a second union of non-dominated solutions and a second Pareto front of non-dominated solutions in the second union by using an improved second multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, wherein the second Pareto front is used to indicate the ability of the improved second multi-objective optimization algorithm MOEA to approach an optimal solution; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0584] a fifth processing unit, configured to obtain a third union of non-dominated solutions and a third Pareto front of non-dominated solutions in the third union by using an improved fourth multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, wherein the third Pareto front is used to indicate the ability of the improved fourth multi-objective optimization algorithm MOEA to approach an optimal solution; the improved fourth multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0585] a sixth processing unit, configured to obtain, by using a preset multi-objective optimization algorithm and the first assembly scheduling test scheme, a fourth union of non-dominated solutions and a fourth Pareto front of the non-dominated solutions in the fourth union, wherein the fourth Pareto front is used to indicate an ability of the preset multi-objective optimization algorithm to approach an optimal solution;

[0586] The second comparison unit is used to compare the first Pareto front, the second Pareto front, the third Pareto front and the fourth Pareto front to obtain a second comparison result indicating the ability of the improved multi-objective optimization algorithm MOEA to approach the optimal solution.

[0587] Optionally, the algorithm improvement module further includes:

[0588] A fourteenth determining unit, configured to obtain a second assembly scheduling test plan according to a preset number of second assembly islands, a preset number of second assembly tasks, and a preset number of second assembly workers;

[0589] a seventh processing unit, configured to obtain a first union of non-dominated solutions through the improved first multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, and obtain a first inverse generation distance analysis result and a first hypervolume index analysis result of the improved first multi-objective optimization algorithm MOEA according to the first union; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved search mechanism strategy;

[0590] an eighth processing unit, configured to obtain a second union of non-dominated solutions through an improved second multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, and to obtain a second inverse generation distance analysis result and a second hypervolume index analysis result of the improved second multi-objective optimization algorithm MOEA according to the second union; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved evolutionary population structure strategy;

[0591] a ninth processing unit, configured to obtain a third union of non-dominated solutions through an improved fourth multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, and to obtain a third inverse generation distance analysis result and a third hypervolume index analysis result of the improved fourth multi-objective optimization algorithm MOEA according to the third union; the improved fourth multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved evolutionary population structure strategy;

[0592] a tenth processing unit, configured to obtain a fourth union of non-dominated solutions by using a preset multi-objective optimization algorithm and the second assembly scheduling test scheme, and to obtain a fourth inverse generation distance analysis result and a fourth hypervolume index analysis result of the preset multi-objective optimization algorithm according to the third union;

[0593] A third comparison unit is used to compare the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result and the fourth inverse generation distance analysis result to obtain a third comparison result indicating the convergence of the improved multi-objective optimization algorithm MOEA;

[0594] The fourth comparison unit is used to compare the first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result to obtain a fourth comparison result indicating the convergence of the improved multi-objective optimization algorithm MOEA.

[0595] Optionally, the algorithm improvement module further includes:

[0596] an eleventh processing unit, configured to obtain a fifth comparison result indicating the stability of the improved multi-objective optimization algorithm MOEA according to the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result, and the fourth inverse generation distance analysis result;

[0597] The twelfth processing unit is used to obtain a sixth comparison result indicating the stability of the improved multi-objective optimization algorithm MOEA based on the first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result.

[0598] Optionally, the algorithm improvement module further includes:

[0599] A thirteenth processing unit is used to obtain a seventh comparison result indicating the dominance of the improved multi-objective optimization algorithm MOEA based on the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result and the fourth inverse generation distance analysis result, taking the improved first multi-objective optimization algorithm MOEA as a benchmark;

[0600] The fourteenth processing unit is used to obtain an eighth comparison result indicating the dominance of the improved multi-objective optimization algorithm MOEA based on the first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result, taking the improved first multi-objective optimization algorithm MOEA as a benchmark.

[0601] Optionally, the algorithm improvement module further includes:

[0602] The fifteenth processing unit is used to obtain a ninth comparison result indicating the applicability of the improved multi-objective optimization algorithm MOEA based on the first Pareto front, the second Pareto front and the third Pareto front.

[0603] Optionally, the algorithm improvement module further includes:

[0604] The sixteenth processing unit is used to obtain a tenth comparison result indicating the applicability of the improved multi-objective optimization algorithm MOEA based on the first inverse generation distance analysis result, the second inverse generation distance analysis result and the third inverse generation distance analysis result.

[0605] Optionally, the algorithm improvement module further includes:

[0606] The seventeenth processing unit is used to obtain an eleventh comparison result indicating the applicability of the improved multi-objective optimization algorithm MOEA based on the first hypervolume index analysis result, the second hypervolume index analysis result and the third hypervolume index analysis result.

[0607] It should be noted that the product assembly scheduling device provided in the embodiment of the present invention is a device capable of executing the above-mentioned product assembly scheduling method. All embodiments of the above-mentioned product assembly scheduling method are applicable to the device and can achieve the same or similar technical effects.

[0608] like Fig.29 As shown, an embodiment of the present invention also provides a product assembly scheduling device, including: a processor 2901, a memory 2902, and a program stored in the memory 2902 and executable on the processor 2901, wherein the program implements the above-mentioned vehicle collision warning method when executed by the processor 2901.

[0609] Optionally, it also includes: a transceiver 2903, and the transceiver 2903 is used to receive and send data under the control of the processor 2901.

[0610] Specifically, the processor 2901 is configured to:

[0611] An integer programming model is established according to the product assembly information; the integer programming model includes a first objective function for indicating the maximum value of minimizing the completion time of the team, a second objective function for indicating minimizing the busy-idle imbalance, and a constraint condition for indicating the assembly workers in the team to perform assembly work;

[0612] According to the product assembly information and the integer programming model, a worker allocation code and a task sequence code are determined by using a double-layer mixed coding method; the worker allocation code is used to indicate the work position that the assembly worker can go to; the task sequence code is used to indicate the order in which the assembly tasks enter the assembly island;

[0613] Based on the integer programming model, a plurality of product assembly scheduling schemes are obtained according to the worker allocation code and the task sequence code;

[0614] By utilizing the improved multi-objective improved algorithm MOEA, the product assembly scheduling optimization scheme is determined among the multiple product assembly scheduling schemes.

[0615] Optionally, the product assembly information includes:

[0616] Assembly island area information;

[0617] Team information in the assembly island;

[0618] All the skills information required for assembly tasks in the assembly island;

[0619] Assembly worker information;

[0620] Assembly task information;

[0621] The coordinate information of the work positions in the team to which the assembly workers can go;

[0622] the skills that assemblers possess and the proficiency of the skills they possess;

[0623] In an assembly task, the basic working hours required for an assembly worker to complete the assembly task corresponding to his or her skills.

[0624] Optionally, the second objective function is determined based on the mean square error of the maximum working time period of each team and the mean square error of the total working time period between teams.

[0625] Optionally, the constraint conditions include:

[0626] The first constraint is used to indicate the skills that the assembly workers have and the proficiency of the skills they have;

[0627] The second constraint is used to indicate whether the assembly worker works in a team;

[0628] The third constraint is used to indicate whether the assembly task is processed on the team;

[0629] The fourth constraint is used to indicate whether the team needs to handle the work tasks corresponding to the skill;

[0630] The fifth constraint is used to indicate the processing order of assembly tasks on the team;

[0631] The sixth constraint is used to indicate the degree to which the proficiency of the assembly workers' skills affects the working hours;

[0632] The seventh constraint is used to indicate the working time of the team in processing the assembly task;

[0633] The eighth constraint condition is used to indicate that the assembly worker can be assigned to a team that requires the target skill to handle the corresponding work task only if the proficiency of the target skill possessed by the assembly worker is not 0;

[0634] The ninth constraint is used to indicate the number of assembly workers assigned to all teams;

[0635] The tenth constraint is used to indicate that the same team on the same assembly island must have the same skills;

[0636] The eleventh constraint is used to indicate the different skills that different teams on the same assembly island must possess;

[0637] The twelfth constraint is used to indicate that each assembly task can only be assigned to one team on each assembly island;

[0638] The thirteenth constraint is used to indicate that the start time of the assembly task on the team is 0, or the end time of the previous assembly task processed by the team;

[0639] The fourteenth constraint condition is used to indicate that the end time of the assembly task on the team is the sum of the start time of the assembly task on the team and the processing time of the assembly task on the team;

[0640] The fifteenth constraint is used to indicate that an assembly task cannot be processed on multiple assembly islands at the same time;

[0641] The sixteenth constraint is used to indicate that a team can only handle one assembly task at the same time.

[0642] Optionally, the processor 2901 is specifically configured to:

[0643] Determine the worker allocation code by integer coding;

[0644] The task sequence code is determined by using an arrangement coding method.

[0645] Optionally, the processor 2901 is specifically configured to:

[0646] Obtaining required position information according to all skill information required for the assembly task in the assembly island in the product assembly information and the team information in the assembly island in the product assembly information;

[0647] The worker allocation code is determined according to the required job information and the skills of the assembly workers while satisfying the constraint conditions.

[0648] Optionally, the processor 2901 is specifically configured to:

[0649] Determining first worker information according to the worker allocation code; the first worker information includes allocation information of assembly workers working on the team;

[0650] Determine first working time information according to the first worker information; the first working time information includes the working time of the assembly workers in each team when processing the assembly task;

[0651] Determine task sequence information according to the task sequence code and the first working time information; the task sequence information includes the order in which each assembly task enters each assembly island;

[0652] Determine the first team information according to the first working hour information and the team work information; the first team information is used to indicate that the target assembly task is processed on the target team; the target assembly task is one of each of the assembly tasks; the target team is one of the teams; the team work information includes: the start time of the team processing the current assembly task, the end time of the team processing the current assembly task and the working time of the team processing the assembly task;

[0653] The plurality of product assembly scheduling plans are obtained according to the task sequence information, the first team information and the first worker information.

[0654] Optionally, the processor 2901 is further specifically configured to:

[0655] Determine a first function value of the first objective function corresponding to each of the product assembly scheduling plans;

[0656] Determine a second function value of the second objective function corresponding to each of the product assembly scheduling plans.

[0657] Optionally, the processor 2901 is specifically configured to:

[0658] Determine, among each of the teams, the first team that has the closest end time for processing the current assembly task as the target team;

[0659] In the case where there are multiple first shifts, among the first shifts, the second shift with the shortest working time is determined as the target shift;

[0660] In the case that there are multiple second teams, the third team in the second teams is determined as the target team; the third team is any one of the second teams.

[0661] Optionally, the processor 2901 is further configured to:

[0662] Through the improvement strategy, the multi-objective improved algorithm MOEA is improved to obtain the improved multi-objective improved algorithm MOEA;

[0663] The improvement strategy includes at least one of the following:

[0664] Improved search mechanism strategies;

[0665] Improved strategies for evolving population structure;

[0666] Improve the population distribution maintenance mechanism strategy.

[0667] Optionally, the processor 2901 is specifically configured to:

[0668] According to a preset mutation mode suitable for permutation coding and a preset mutation mode suitable for integer coding, a mutation combination complex scheme is obtained;

[0669] According to the mutation combination scheme, the evolved offspring are further searched to obtain the mutant offspring, so that the improved multi-objective improved algorithm MOEA searches among the evolved offspring, the mutant offspring and the population parent generation.

[0670] Optionally, the processor 2901 is specifically configured to:

[0671] Divide the evolutionary offspring into non-dominant hierarchies;

[0672] Determine the evolutionary offspring whose non-dominant level is less than or equal to the first preset level as the elite offspring;

[0673] The evolutionary offspring are searched according to the elite offspring to obtain the mutant elite offspring, so that the improved multi-objective improved algorithm MOEA searches in the mutant elite offspring.

[0674] Optionally, the processor 2901 is specifically configured to:

[0675] Divide the evolutionary offspring into non-dominant hierarchies;

[0676] According to the non-kinship calculation function, the crowding distance strategy is improved so that the improved multi-objective improved algorithm MOEA generates the next generation of individuals according to the evolutionary offspring of the target non-dominated hierarchy;

[0677] Wherein, the non-kinship calculation function is determined based on the crowding distance and the similarity;

[0678] The target non-dominated hierarchy is a non-dominated hierarchy in which the total number of individuals is less than the preset number of individuals in the population, and the total number of individuals after merging is greater than the preset number of individuals in the population.

[0679] Optionally, the processor 2901 is further configured to:

[0680] Analyze the complexity of the improved first multi-objective improved algorithm MOEA to obtain a first complexity analysis result; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved search mechanism strategy;

[0681] Analyze the complexity of the improved second multi-objective improved algorithm MOEA to obtain a second complexity analysis result; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved evolutionary population structure strategy;

[0682] Analyze the complexity of the improved third multi-objective improved algorithm MOEA to obtain a third complexity analysis result; the improved third multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved population distribution maintaining mechanism strategy;

[0683] Analyze the complexity of the preset multi-objective improved algorithm to obtain a fourth complexity analysis result;

[0684] The first complexity analysis result, the second complexity analysis result, the third complexity analysis result and the fourth complexity analysis result are compared to obtain a first comparison result indicating the complexity of the improved multi-objective improved algorithm MOEA.

[0685] Optionally, the processor 2901 is further configured to:

[0686] Obtaining a first assembly scheduling test plan according to a preset number of first assembly islands, a preset number of first assembly tasks, and a preset number of first assembly workers;

[0687] By using the improved first multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, a first union of non-dominated solutions and a first Pareto front of non-dominated solutions in the first union are obtained, wherein the first Pareto front is used to indicate the ability of the improved first multi-objective optimization algorithm MOEA to approach the optimal solution; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved search mechanism strategy;

[0688] By using the improved second multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, a second union of non-dominated solutions and a second Pareto front of non-dominated solutions in the second union are obtained, wherein the second Pareto front is used to indicate the ability of the improved second multi-objective optimization algorithm MOEA to approach the optimal solution; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0689] By using the improved fourth multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, a third union of non-dominated solutions and a third Pareto front of non-dominated solutions in the third union are obtained, wherein the third Pareto front is used to indicate the ability of the improved fourth multi-objective optimization algorithm MOEA to approach the optimal solution; the improved fourth multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0690] Obtaining, by using a preset multi-objective optimization algorithm and the first assembly scheduling test scheme, a fourth union of non-dominated solutions and a fourth Pareto front of the non-dominated solutions in the fourth union, wherein the fourth Pareto front is used to indicate the ability of the preset multi-objective optimization algorithm to approach an optimal solution;

[0691] The first Pareto front, the second Pareto front, the third Pareto front and the fourth Pareto front are compared to obtain a second comparison result indicating the ability of the improved multi-objective optimization algorithm MOEA to approach the optimal solution.

[0692] Optionally, the processor 2901 is further configured to:

[0693] Obtaining a second assembly scheduling test plan according to a preset number of second assembly islands, a preset number of second assembly tasks, and a preset number of second assembly workers;

[0694] By using the improved first multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, a first union of non-dominated solutions is obtained, and according to the first union, a first inverse generation distance analysis result and a first hypervolume index analysis result of the improved first multi-objective optimization algorithm MOEA are obtained; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved search mechanism strategy;

[0695] By using the improved second multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, a second union of non-dominated solutions is obtained, and according to the second union, a second inverse generation distance analysis result and a second hypervolume index analysis result of the improved second multi-objective optimization algorithm MOEA are obtained; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0696] By using the improved fourth multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, a third union of non-dominated solutions is obtained, and according to the third union, a third inverse generation distance analysis result and a third hypervolume index analysis result of the improved fourth multi-objective optimization algorithm MOEA are obtained; the improved fourth multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy;

[0697] Obtaining a fourth union of non-dominated solutions through a preset multi-objective optimization algorithm and the second assembly scheduling test scheme, and obtaining a fourth inverse generation distance analysis result and a fourth hypervolume index analysis result of the preset multi-objective optimization algorithm according to the third union;

[0698] Comparing the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result and the fourth inverse generation distance analysis result to obtain a third comparison result indicating the convergence of the improved multi-objective optimization algorithm MOEA;

[0699] The first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result are compared to obtain a fourth comparison result indicating the convergence of the improved multi-objective optimization algorithm MOEA.

[0700] Optionally, the processor 2901 is further configured to:

[0701] Obtaining a fifth comparison result indicating the stability of the improved multi-objective optimization algorithm MOEA according to the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result, and the fourth inverse generation distance analysis result;

[0702] According to the first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result, a sixth comparison result indicating the stability of the improved multi-objective optimization algorithm MOEA is obtained.

[0703] Optionally, the processor 2901 is further configured to:

[0704] Taking the improved first multi-objective optimization algorithm MOEA as a benchmark, obtaining a seventh comparison result indicating the dominance of the improved multi-objective optimization algorithm MOEA according to the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result and the fourth inverse generation distance analysis result;

[0705] Taking the improved first multi-objective optimization algorithm MOEA as a benchmark, according to the first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result, an eighth comparison result indicating the dominance of the improved multi-objective optimization algorithm MOEA is obtained.

[0706] Optionally, the processor 2901 is further configured to:

[0707] According to the first Pareto front, the second Pareto front and the third Pareto front, a ninth comparison result indicating the applicability of the improved multi-objective optimization algorithm MOEA is obtained.

[0708] Optionally, the processor 2901 is further configured to:

[0709] According to the first inverse generation distance analysis result, the second inverse generation distance analysis result and the third inverse generation distance analysis result, a tenth comparison result indicating the applicability of the improved multi-objective optimization algorithm MOEA is obtained.

[0710] Optionally, the processor 2901 is further configured to:

[0711] According to the first hypervolume index analysis result, the second hypervolume index analysis result and the third hypervolume index analysis result, an eleventh comparison result is obtained for indicating the applicability of the improved multi-objective optimization algorithm MOEA.

[0712] Among them, Fig.29 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically linking together various circuits of one or more processors represented by processor 2901 and memory represented by memory 2902. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides a user interface 2904. The transceiver 2903 may be a plurality of components, i.e., including a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium. The processor 2901 is responsible for managing the bus architecture and general processing, and the memory 2902 may store data used by the processor 2901 when performing operations.

[0713] In addition, a specific embodiment of the present invention further provides a readable storage medium on which a computer program is stored, wherein when the program is executed by a processor, the steps in the product assembly scheduling method as described above are implemented.

[0714] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary personnel in this technical field, several improvements and modifications can be made without departing from the principles described in the present invention. These improvements and modifications are also within the scope of protection of the present invention.

Claims

1. A product assembly scheduling method, characterized in that: include: An integer programming model is established according to the product assembly information; the integer programming model includes a first objective function for indicating the maximum value of minimizing the completion time of the team, a second objective function for indicating minimizing the busy-idle imbalance, and a constraint condition for indicating the assembly workers in the team to perform assembly work; According to the product assembly information and the integer programming model, a worker allocation code and a task sequence code are determined by using a double-layer mixed coding method; The worker allocation code is used to indicate the work station that the assembly worker can go to; the task sequence code is used to indicate the order in which the assembly tasks enter the assembly island; Based on the integer programming model, a plurality of product assembly scheduling schemes are obtained according to the worker allocation code and the task sequence code; Determine the product assembly scheduling optimization scheme among the multiple product assembly scheduling schemes by using the improved multi-objective improved algorithm MOEA; Wherein, the product assembly information includes: Assembly island area information; Team information in the assembly island; All the skills information required for assembly tasks in the assembly island; Assembly worker information; Assembly task information; The coordinate information of the work positions in the team to which the assembly workers can go; the skills that assemblers possess and the proficiency of the skills they possess; In assembly tasks, the basic working hours required for assembly workers to complete the assembly tasks corresponding to their skills; The constraints include: The first constraint is used to indicate the skills that the assembly workers have and the proficiency of the skills they have; The second constraint is used to indicate whether the assembly worker works in a team; The third constraint is used to indicate whether the assembly task is processed on the team; The fourth constraint is used to indicate whether the team needs to handle the work tasks corresponding to the skill; The fifth constraint is used to indicate the processing order of assembly tasks on the team; The sixth constraint is used to indicate the degree to which the proficiency of the assembly workers' skills affects the working hours; The seventh constraint is used to indicate the working time of the team in processing the assembly task; The eighth constraint condition is used to indicate that the assembly worker can be assigned to a team that requires the target skill to handle the corresponding work task only if the proficiency of the target skill possessed by the assembly worker is not 0; The ninth constraint is used to indicate the number of assembly workers assigned to all teams; The tenth constraint is used to indicate that the same team on the same assembly island must have the same skills; The eleventh constraint is used to indicate the different skills that different teams on the same assembly island must possess; The twelfth constraint is used to indicate that each assembly task can only be assigned to one team on each assembly island; The thirteenth constraint is used to indicate that the start time of the assembly task on the team is 0, or the end time of the previous assembly task processed by the team; The fourteenth constraint condition is used to indicate that the end time of the assembly task on the team is the sum of the start time of the assembly task on the team and the processing time of the assembly task on the team; The fifteenth constraint is used to indicate that an assembly task cannot be processed on multiple assembly islands at the same time; The sixteenth constraint is used to indicate that a team can only handle one assembly task at the same time.

2. The product assembly scheduling method according to claim 1, characterized in that: The second objective function is determined based on the mean square error of the maximum working time period of each team and the mean square error of the total working time period between teams.

3. The product assembly scheduling method according to claim 1, characterized in that: A two-layer hybrid coding method is used to determine the worker allocation code and task sequence code, including: Determine the worker allocation code by integer coding; The task sequence code is determined by using an arrangement coding method.

4. The product assembly scheduling method according to claim 1, characterized in that: According to the product assembly information and the integer programming model, a worker allocation code is determined using a double-layer hybrid coding method, including: Obtaining required position information according to all skill information required for the assembly task in the assembly island in the product assembly information and the team information in the assembly island in the product assembly information; The worker allocation code is determined according to the required job information and the skills of the assembly workers while satisfying the constraint conditions.

5. The product assembly scheduling method according to claim 1, characterized in that: Based on the integer programming model, multiple product assembly scheduling solutions are obtained according to the worker allocation code and the task sequence code, including: Determining first worker information according to the worker allocation code; the first worker information includes allocation information of assembly workers working on the team; Determine first working time information according to the first worker information; the first working time information includes the working time of the assembly workers in each team when processing the assembly task; Determine task sequence information according to the task sequence code and the first working time information; the task sequence information includes the order in which each assembly task enters each assembly island; Determine the first team information according to the first working hour information and the team work information; the first team information is used to indicate that the target assembly task is processed on the target team; the target assembly task is one of each of the assembly tasks; the target team is one of the teams; the team work information includes: the start time of the team processing the current assembly task, the end time of the team processing the current assembly task and the working time of the team processing the assembly task; The plurality of product assembly scheduling plans are obtained according to the task sequence information, the first team information and the first worker information.

6. The product assembly scheduling method according to claim 5, characterized in that: The method further comprises: Determine a first function value of the first objective function corresponding to each of the product assembly scheduling plans; Determine a second function value of the second objective function corresponding to each of the product assembly scheduling plans.

7. The product assembly scheduling method according to claim 5, characterized in that: Determining first team information according to the first working hour information and team work information includes: Determine, among each of the teams, the first team that has the closest end time for processing the current assembly task as the target team; In the case where there are multiple first shifts, determining the second shift with the shortest working time among the first shifts as the target shift; In the case that there are multiple second teams, the third team in the second teams is determined as the target team; the third team is any one of the second teams.

8. The product assembly scheduling method according to claim 1, characterized in that: According to the improved multi-objective improved algorithm MOEA, before the multiple product assembly scheduling schemes determine the product assembly scheduling optimization scheme, the method further includes: Through the improvement strategy, the multi-objective improved algorithm MOEA is improved to obtain the improved multi-objective improved algorithm MOEA; The improvement strategy includes at least one of the following: Improved search mechanism strategies; Improved strategies for evolving population structure; Improve the population distribution maintenance mechanism strategy.

9. The product assembly scheduling method according to claim 8, characterized in that: By improving the search mechanism strategy, the multi-objective improved algorithm MOEA is improved to obtain an improved multi-objective improved algorithm MOEA, including: According to a preset mutation mode suitable for permutation coding and a preset mutation mode suitable for integer coding, a mutation combination complex scheme is obtained; According to the mutation combination complex scheme, the evolved offspring are further searched to obtain the mutant offspring, so that the improved multi-objective improved algorithm MOEA searches among the evolved offspring, the mutant offspring and the population parent generation.

10. The product assembly scheduling method according to claim 9, characterized in that: By improving the evolutionary population structure strategy, the multi-objective improved algorithm MOEA is improved to obtain an improved multi-objective improved algorithm MOEA, including: Divide the evolutionary offspring into non-dominant hierarchies; Determine the evolutionary offspring whose non-dominant level is less than or equal to the first preset level as the elite offspring; The evolutionary offspring are searched according to the elite offspring to obtain the mutant elite offspring, so that the improved multi-objective improved algorithm MOEA searches in the mutant elite offspring.

11. The product assembly scheduling method according to claim 8, characterized in that: By improving the population distribution maintenance mechanism strategy, the multi-objective improved algorithm MOEA is improved to obtain an improved multi-objective improved algorithm MOEA, including: Divide the evolutionary offspring into non-dominant hierarchies; According to the non-kinship calculation function, the crowding distance strategy is improved so that the improved multi-objective improved algorithm MOEA generates the next generation of individuals according to the evolutionary offspring of the target non-dominated hierarchy; Wherein, the non-kinship calculation function is determined based on the crowding distance and the similarity; The target non-dominated hierarchy is a non-dominated hierarchy in which the total number of individuals is less than the preset number of individuals in the population, and the total number of individuals after merging is greater than the preset number of individuals in the population.

12. The product assembly scheduling method according to claim 8, characterized in that: The method further comprises: Analyze the complexity of the improved first multi-objective improved algorithm MOEA to obtain a first complexity analysis result; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved search mechanism strategy; Analyze the complexity of the improved second multi-objective improved algorithm MOEA to obtain a second complexity analysis result; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved evolutionary population structure strategy; Analyze the complexity of the improved third multi-objective improved algorithm MOEA to obtain a third complexity analysis result; the improved third multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA through the improved population distribution maintaining mechanism strategy; Analyze the complexity of the preset multi-objective improved algorithm to obtain a fourth complexity analysis result; The first complexity analysis result, the second complexity analysis result, the third complexity analysis result and the fourth complexity analysis result are compared to obtain a first comparison result indicating the complexity of the improved multi-objective improved algorithm MOEA.

13. The product assembly scheduling method according to claim 8, characterized in that: The method further comprises: Obtaining a first assembly scheduling test plan according to a preset number of first assembly islands, a preset number of first assembly tasks, and a preset number of first assembly workers; By using the improved first multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, a first union of non-dominated solutions and a first Pareto front of non-dominated solutions in the first union are obtained, wherein the first Pareto front is used to indicate the ability of the improved first multi-objective optimization algorithm MOEA to approach the optimal solution; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved search mechanism strategy; By using the improved second multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, a second union of non-dominated solutions and a second Pareto front of non-dominated solutions in the second union are obtained, wherein the second Pareto front is used to indicate the ability of the improved second multi-objective optimization algorithm MOEA to approach the optimal solution; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy; By using the improved fourth multi-objective optimization algorithm MOEA and the first assembly scheduling test scheme, a third union of non-dominated solutions and a third Pareto front of non-dominated solutions in the third union are obtained, wherein the third Pareto front is used to indicate the ability of the improved fourth multi-objective optimization algorithm MOEA to approach the optimal solution; the improved fourth multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy; Obtaining, by using a preset multi-objective optimization algorithm and the first assembly scheduling test scheme, a fourth union of non-dominated solutions and a fourth Pareto front of the non-dominated solutions in the fourth union, wherein the fourth Pareto front is used to indicate the ability of the preset multi-objective optimization algorithm to approach an optimal solution; The first Pareto front, the second Pareto front, the third Pareto front and the fourth Pareto front are compared to obtain a second comparison result indicating the ability of the improved multi-objective optimization algorithm MOEA to approach the optimal solution.

14. The product assembly scheduling method according to claim 8, characterized in that: The method further comprises: Obtaining a second assembly scheduling test plan according to a preset number of second assembly islands, a preset number of second assembly tasks, and a preset number of second assembly workers; By using the improved first multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, a first union of non-dominated solutions is obtained, and according to the first union, a first inverse generation distance analysis result and a first hypervolume index analysis result of the improved first multi-objective optimization algorithm MOEA are obtained; the improved first multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved search mechanism strategy; By using the improved second multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, a second union of non-dominated solutions is obtained, and according to the second union, a second inverse generation distance analysis result and a second hypervolume index analysis result of the improved second multi-objective optimization algorithm MOEA are obtained; the improved second multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy; By using the improved fourth multi-objective optimization algorithm MOEA and the second assembly scheduling test scheme, a third union of non-dominated solutions is obtained, and according to the third union, a third inverse generation distance analysis result and a third hypervolume index analysis result of the improved fourth multi-objective optimization algorithm MOEA are obtained; the improved fourth multi-objective improved algorithm MOEA is obtained by improving the multi-objective improved algorithm MOEA by using the improved evolutionary population structure strategy; Obtaining a fourth union of non-dominated solutions through a preset multi-objective optimization algorithm and the second assembly scheduling test scheme, and obtaining a fourth inverse generation distance analysis result and a fourth hypervolume index analysis result of the preset multi-objective optimization algorithm according to the third union; Comparing the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result and the fourth inverse generation distance analysis result to obtain a third comparison result indicating the convergence of the improved multi-objective optimization algorithm MOEA; The first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result are compared to obtain a fourth comparison result indicating the convergence of the improved multi-objective optimization algorithm MOEA.

15. The product assembly scheduling method according to claim 14, characterized in that: The method further comprises: Obtaining a fifth comparison result indicating the stability of the improved multi-objective optimization algorithm MOEA according to the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result, and the fourth inverse generation distance analysis result; According to the first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result, a sixth comparison result indicating the stability of the improved multi-objective optimization algorithm MOEA is obtained.

16. The product assembly scheduling method according to claim 14, characterized in that: The method further comprises: Taking the improved first multi-objective optimization algorithm MOEA as a benchmark, obtaining a seventh comparison result indicating the dominance of the improved multi-objective optimization algorithm MOEA according to the first inverse generation distance analysis result, the second inverse generation distance analysis result, the third inverse generation distance analysis result and the fourth inverse generation distance analysis result; Taking the improved first multi-objective optimization algorithm MOEA as a benchmark, according to the first hypervolume index analysis result, the second hypervolume index analysis result, the third hypervolume index analysis result and the fourth hypervolume index analysis result, an eighth comparison result indicating the dominance of the improved multi-objective optimization algorithm MOEA is obtained.

17. The product assembly scheduling method according to claim 13, characterized in that: The method further comprises: According to the first Pareto front, the second Pareto front and the third Pareto front, a ninth comparison result indicating the applicability of the improved multi-objective optimization algorithm MOEA is obtained.

18. The product assembly scheduling method according to claim 14, characterized in that: The method further comprises: According to the first inverse generation distance analysis result, the second inverse generation distance analysis result and the third inverse generation distance analysis result, a tenth comparison result indicating the applicability of the improved multi-objective optimization algorithm MOEA is obtained.

19. The product assembly scheduling method according to claim 14, characterized in that: The method further comprises: According to the first hypervolume index analysis result, the second hypervolume index analysis result and the third hypervolume index analysis result, an eleventh comparison result is obtained for indicating the applicability of the improved multi-objective optimization algorithm MOEA.

20. A product assembly scheduling device, characterized in that: include: A model building module, used to build an integer programming model according to product assembly information; the integer programming model includes a first objective function for indicating the maximum value of minimizing the completion time of the team, a second objective function for indicating minimizing the busy-idle imbalance, and a constraint condition for indicating the assembly workers in the team to perform assembly work; A first determination module is used to determine a worker allocation code and a task sequence code according to the product assembly information and the integer programming model using a double-layer hybrid coding method; The worker allocation code is used to indicate the work station that the assembly worker can go to; the task sequence code is used to indicate the order in which the assembly tasks enter the assembly island; A second determination module is used to obtain a plurality of product assembly scheduling solutions based on the integer programming model and according to the worker allocation code and the task sequence code; A third determination module is used to determine a product assembly scheduling optimization scheme from the multiple product assembly scheduling schemes using an improved multi-objective improved algorithm MOEA; Wherein, the product assembly information includes: Assembly island area information; Team information in the assembly island; All the skills information required for assembly tasks in the assembly island; Assembly worker information; Assembly task information; The coordinate information of the work positions in the team to which the assembly workers can go; the skills that assemblers possess and the proficiency of the skills they possess; In assembly tasks, the basic working hours required for assembly workers to complete the assembly tasks corresponding to their skills; The constraints include: The first constraint is used to indicate the skills that the assembly workers have and the proficiency of the skills they have; The second constraint is used to indicate whether the assembly worker works in a team; The third constraint is used to indicate whether the assembly task is processed on the team; The fourth constraint is used to indicate whether the team needs to handle the work tasks corresponding to the skill; The fifth constraint is used to indicate the processing order of assembly tasks on the team; The sixth constraint is used to indicate the degree to which the proficiency of the assembly workers' skills affects the working hours; The seventh constraint is used to indicate the working time of the team in processing the assembly task; The eighth constraint condition is used to indicate that the assembly worker can be assigned to a team that requires the target skill to handle the corresponding work task only if the proficiency of the target skill possessed by the assembly worker is not 0; The ninth constraint is used to indicate the number of assembly workers assigned to all teams; The tenth constraint is used to indicate that the same team on the same assembly island must have the same skills; The eleventh constraint is used to indicate the different skills that different teams on the same assembly island must possess; The twelfth constraint is used to indicate that each assembly task can only be assigned to one team on each assembly island; The thirteenth constraint is used to indicate that the start time of the assembly task on the team is 0, or the end time of the previous assembly task processed by the team; The fourteenth constraint condition is used to indicate that the end time of the assembly task on the team is the sum of the start time of the assembly task on the team and the processing time of the assembly task on the team; The fifteenth constraint is used to indicate that an assembly task cannot be processed on multiple assembly islands at the same time; The sixteenth constraint is used to indicate that a team can only handle one assembly task at the same time.

21. A product assembly scheduling device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the steps of the product assembly scheduling method according to any one of claims 1 to 19 are implemented.

22. A readable storage medium, characterized in that: The readable storage medium stores a program, and when the program is executed by a processor, the steps in the product assembly scheduling method according to any one of claims 1 to 19 are implemented.