Intelligent optimization disassembly line balancing method based on CTGAN and improved NSGA-II algorithm

Abstract

The present invention belongs to the field of intelligent optimization algorithms and disassembly line balance, and relates to an intelligent optimization disassembly line balance method based on CTGAN and an improved NSGA-II algorithm. Based on historical disassembly data, this method uses a generative AI method CTGAN that expands table data to expand the disassembly data of products to be disassembled, which accounts for a relatively small proportion, and establishes a disassembly line balance solution model for mixed homogeneous products under a fine-grained mixed disassembly mode. The improved intelligent optimization method NSGA-II algorithm is used for solution, which greatly improves the disassembly efficiency and reduces the disassembly cost and disassembly energy consumption.

Description

Technical Field

[0001] The invention belongs to the field of intelligent optimization algorithm and disassembly line balance, and relates to an intelligent optimization disassembly line balance method based on CTGAN and an improved NSGA-II algorithm. Background Art

[0002] In the actual operation of the disassembly line, there are different degrees of differences in the types, models, and degree of wear of the products to be disassembled. Considering the disassembly cost, disassembly companies usually do not classify the products to be disassembled in a more detailed manner, but directly disassemble the mixed products to be disassembled of the same type on the same disassembly line; however, in the actual disassembly process, the proportion of mixed homogeneous products to be disassembled is unbalanced, and there is a data imbalance problem. The existing solution to the balance problem of mixed disassembly lines usually adopts the method of averaging to comprehensively consider the mixed homogeneous products to be disassembled. For a large number of products to be disassembled, the amount of data is large, and the result obtained by averaging can better eliminate the influence of abnormal data, thereby better describing the disassembly of the product.

[0003] For a relatively small number of products to be disassembled, the existing solution to the disassembly line balance problem does not expand the disassembly data of the relatively small number of products to be disassembled, but directly averages them, and uses the average value to directly characterize the disassembly situation of the product to be disassembled, and then uses the average value as the input of the model for solution. Especially in the disassembly line in the mixed disassembly scenario, mixed homogeneous products to be disassembled are disassembled on the same disassembly line. In the scenario where the proportion of different products to be disassembled in all products to be disassembled is not balanced, if the unbalanced disassembly data is not processed, it is likely to lead to low disassembly efficiency or even disassembly failure. Therefore, data expansion of the disassembly data of products to be disassembled with a small number and a low proportion is a key factor in optimizing the disassembly line balance problem.

[0004] In addition, in the actual disassembly process, a single destructive disassembly mode (workers usually use destructive disassembly modes such as direct shearing, sawing, crushing, etc. to achieve the purpose of rapid disassembly) or a single non-destructive disassembly mode does not meet the disassembly requirements. Among them, parts that are hazardous or have recycling needs must be disassembled using a non-destructive non-destructive disassembly mode. For parts that are non-hazardous and have no recycling needs, a destructive disassembly mode that directly destroys the part structure is usually adopted to achieve the purpose of rapid disassembly. However, existing research is limited to the optimization solution of the disassembly line balance problem on the mixed disassembly line under the mixed disassembly mode, and has not refined the degree of hazard and demand, which leads to the problem of not being able to fully utilize resources, improve disassembly efficiency, and reduce disassembly energy consumption.

[0005] Therefore, a fine-grained classification of the harmfulness and demand of the disassembly tasks of the products to be disassembled is carried out, and a disassembly line balance model of mixed homogeneous products under a fine-grained mixed disassembly mode is established. The improved intelligent optimization method is used to solve the problem, which can effectively solve the problems of insufficient resource utilization, low disassembly efficiency and high disassembly energy consumption in the disassembly line balance problem. Summary of the invention

[0006] In order to solve the problem of low disassembly efficiency caused by the unbalanced proportion of products to be disassembled in the mixed disassembly line and the problem of insufficient granularity of classification of the hazards and demand of the tasks to be disassembled during the disassembly process, the present invention provides an intelligent optimization disassembly line balancing method based on CTGAN and an improved NSGA-II algorithm. On the basis of historical disassembly data, a generative AI method CTGAN that expands tabular data is used to predict the disassembly data of products to be disassembled with a smaller disassembly ratio, and a disassembly line balancing model for mixed homogeneous products under a fine-grained mixed disassembly mode is established. The improved NSGA-II method is used to solve the model, and the disassembly sequence of mixed homogeneous products to be disassembled on the mixed disassembly line is optimized to improve the disassembly efficiency.

[0007] The present invention is achieved through the following technical solutions:

[0008] An intelligent optimization disassembly line balancing method based on CTGAN and an improved NSGA-II algorithm, the method comprising:

[0009] (1) Collect historical disassembly data;

[0010] (2) In the currently processed batch of orders, when the product type of the product to be disassembled accounts for less than 1 / 3 of the product type in the historical disassembly data, CTGAN is used to expand the disassembly data of the product to be disassembled to obtain the expanded disassembly data;

[0011] (3) Using the expanded disassembly data, fine-grained classification is performed on the harmfulness and demand of the disassembly tasks of the products to be disassembled, and a disassembly line balance solution model for mixed homogeneous products under a fine-grained mixed disassembly mode is established;

[0012] (4) Using the improved NSGA-II algorithm to solve the disassembly line balance solution model established in step (3) to obtain a solution to the current objective function; and by decoding the solution, a disassembly planning sequence is obtained.

[0013] Furthermore, in step (1), the historical disassembly data includes order number, product type, product model, product quantity, disassembly time of different disassembly modes, disassembly energy consumption of different disassembly modes, disassembly cost of different disassembly modes, disassembly priority relationship constraint diagram of products, and disassembly feasible sequence of products.

[0014] Furthermore, step (1) also includes preprocessing the historical disassembly data; the preprocessing includes the steps of: cleaning data, outlier processing, filling missing values, standardization and normalization, and feature engineering.

[0015] Furthermore, in step (3), the disassembly line balance solution model for mixed homogeneous products under the fine-grained mixed disassembly mode includes: minimizing the number of workstations objective function , minimizing the workstation smoothing exponential objective function , minimize the cost objective function , minimize the energy consumption objective function , minimize the pollution index objective function .

[0016] Furthermore, the objective function of minimizing the number of workstations is The formula is:

[0017] ;

[0018] Where m is the workstation index, M is the total number of workstations, m=1,...,M; Indicates whether the mth workstation is turned on. When the workstation is turned on is equal to 1, when the workstation is shut down, is equal to 0.

[0019] Furthermore, the workstation smoothness index objective function is minimized The formula is:

[0020] ;

[0021] In the formula, i，l represents the disassembly task index, I represents the total number of disassembly tasks, , , ; j represents the position in the disassembly sequence, and J represents the total number of the largest disassembly tasks; Indicates the actual cycle time; Indicates whether task i is assigned to workstation m. When task i is assigned to workstation m is equal to 1 when task i is not assigned to workstation m is equal to 0; Representation Task l Whether to assign it to workstation m, when the task l When assigned to workstation m Equal to 1, when the task l When no workstation m is assigned is equal to 0; Indicates whether task i is in the jth position of workstation m. When task i is in the jth position of workstation m, Equal to 1, when task i is not in the jth position of workstation m =0; when task i is in the j+1th position of workstation m Equal to 1, when task i is not in the j+1th position of workstation m is equal to 0; Indicates whether there is a tool switch between task i and task l. When there is a tool switch between task i and task l Equal to 1, when there is no tool switching between task i and task l is equal to 0; The time required for a tool switch to occur;

[0022] represents the average disassembly time of task i, , K is the number of product types, k represents the product type, Indicates the disassembly ratio of k-type products; Indicates whether there is task i in product type k. If there is task i in product type k Equal to 1, when there is no task i in the k-type product is equal to 0; represents the disassembly time of product type k in disassembly task i; , , , represents the disassembly time of product type k in disassembly task i using non-destructive disassembly mode, represents the disassembly time of product type k in disassembly task i using destructive disassembly mode, Indicates the disassembly mode adopted by task i. When task i adopts the destructive mode Equal to 1, when task i adopts non-destructive mode is equal to 0.

[0023] Furthermore, minimizing the cost objective function The formula is:

[0024] ;

[0025] In the formula, Represents the cost per unit time of a workstation; It represents the additional cost per unit time for a workstation to process the primary hazardous product; It represents the additional cost per unit time for a workstation to process secondary harmful products; It represents the additional cost per unit time for a workstation to process the third-level harmful products; Indicates whether workstation m has a first-level dangerous task. If workstation m has a first-level dangerous task, then is 1, otherwise is 0; Indicates whether workstation m has a second-level dangerous task. If workstation m has a second-level dangerous task, then is 1, otherwise is 0; Indicates whether workstation m has a third-level dangerous task. If workstation m has a third-level dangerous task, then is 1, otherwise is 0;

[0026] represents the average disassembly cost of task i, , , represents the disassembly cost of product type k in disassembly task i, , , , is the disassembly cost of product type k in disassembly task i using non-destructive disassembly mode, is the disassembly cost of product type k in disassembly task i using destructive disassembly mode; Indicates the disassembly mode adopted by task i. When task i adopts the destructive mode Equal to 1, when task i adopts non-destructive mode is equal to 0.

[0027] Furthermore, the energy consumption objective function is minimized The formula is:

[0028] ;

[0029] In the formula, It represents the energy consumption per unit time of a workstation; It represents the additional energy consumption per unit time of a workstation in processing primary harmful products; It represents the additional energy consumption per unit time of a workstation in processing secondary harmful products; It represents the additional energy consumption per unit time of a workstation in processing the third-level harmful products;

[0030] represents the average disassembly energy consumption of task i, , , represents the disassembly energy consumption of product type k in disassembly task i; , , , represents the disassembly energy consumption of product type k in disassembly task i using non-destructive disassembly mode, It represents the disassembly energy consumption of product type k in disassembly task i using destructive disassembly mode.

[0031] Furthermore, the pollution index objective function min is minimized f The formula for 5 is:

[0032] ;

[0033] In the formula, represents the average pollution index of task i; , , represents the pollution index of type k product in disassembly task i, , , , represents the contamination index of product type k in non-destructive disassembly mode in disassembly task i, It represents the pollution index of product type k in the destructive disassembly mode of disassembly task i.

[0034] Furthermore, step (4) specifically includes the following steps:

[0035] (4.1) Coding: The coding method of NSGA-II is improved, and a matrix of 2 rows and 1 columns is used to encode the solution. The first row is the task number to be disassembled, and the second row is the task disassembly mode. The disassembly mode includes destructive disassembly mode and non-destructive disassembly mode. 0 represents non-destructive mode, 1 represents destructive mode, and the number of columns I is the total number of disassembly tasks.

[0036] (4.2) Initialization: In the initial solution generation, the tasks to be disassembled are arranged in ascending order according to the number of times they are selected. If the number of times they are selected is the same, they are arranged in descending order according to the hazard level, and the first task after sorting is selected;

[0037] (4.3) Non-dominated sorting and crowding calculation: Perform non-dominated sorting on the current solution, calculate crowding, sort in descending order according to crowding, and select the individuals to enter the next cycle;

[0038] (4.4) Crossover: Use the two-point crossover method to randomly generate two crossover points, position1 and position2. The part between the crossover points position1 and position2 is the part to be crossed. Generate offspring chromosomes that meet the priority relationship by exchanging the same gene sequences of the two parts to be crossed.

[0039] (4.5) Mutation: Perform mutation operations on the solution, using a two-stage mutation method to ensure that the generated offspring can meet the constraints; in the first stage, the location of the mutation point is randomly selected. If a task is harmful or has a recycling requirement, the disassembly mode of the corresponding task will not be changed. Otherwise, whether to change the disassembly mode of the corresponding task is determined based on the randomly generated probability; in the second stage, the direct predecessor and direct successor of the task are found, and a position is selected between the direct predecessor and the direct successor, and the corresponding task is randomly inserted;

[0040] (4.6) Loop: Perform non-dominated sorting and crowding calculation on the current population, perform non-dominated sorting on the current solution, calculate crowding, and select individuals that can enter the next loop; until the termination condition is reached, the loop ends and the solution under the current objective function and constraint conditions is obtained;

[0041] (4.7) Decoding: Arrange the tasks to be disassembled to the workstations in the order on the chromosome. If the current workstation is full, open the next workstation, and so on, until all the disassembly tasks are arranged.

[0042] Beneficial technical effects of the present invention:

[0043] The intelligent optimization disassembly line balancing method based on CTGAN and improved NSGA-II algorithm proposed in the present invention is an intelligent optimization disassembly line balancing method for mixed homogeneous products under a fine-grained mixed disassembly mode based on generative AI and NSGA-II. The method can be applied to the field of intelligent disassembly; on the basis of historical disassembly data, the method uses a generative AI method CTGAN for expanding tabular data to expand the disassembly data of products to be disassembled which accounts for a relatively small proportion, establishes a disassembly line balance solution model for mixed homogeneous products under a fine-grained mixed disassembly mode, and uses the improved intelligent optimization method NSGA-II algorithm for solving it, which greatly improves the disassembly efficiency and reduces the disassembly cost and disassembly energy consumption.

[0044] The intelligent optimization disassembly line balancing method proposed in the present invention classifies the hazard level and demand level of the disassembly tasks in a fine-grained manner, wherein hazardous or demanding components must adopt a non-destructive disassembly mode, and the remaining components can randomly choose a destructive or non-destructive disassembly mode. On this basis, a disassembly line balancing model for mixed homogeneous products under a fine-grained mixed disassembly mode is established; this model can better fit the actual disassembly process, and by solving the model, a disassembly sequence planning scheme that better meets actual needs can be obtained.

[0045] The intelligent optimization disassembly line balancing method proposed in the present invention adopts the improved multi-objective solution method NSGA-II algorithm to solve the model. By improving the encoding strategy, decoding strategy, initialization method, crossover operation, mutation operation and other parts of the NSGA-II algorithm, the solution to the problem can be better searched according to the actual situation of the problem. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a flow chart of an intelligent optimization disassembly line balancing method based on CTGAN and improved NSGA-II algorithm in an embodiment of the present invention;

[0047] Figure 2 The chromosome encoding method in the embodiment of the present invention;

[0048] Figure 3-1 The crossover generation of offspring 1 in the embodiment of the present invention is represented by the parent chromosome

[0049] Figure 3-2 For the crossover generation of offspring 1 in the embodiment of the present invention - parent chromosome crossover point selection

[0050] Figure 3-3 Crossover to generate offspring 1 in the embodiment of the present invention - crossover of parent chromosome fragments

[0051] Figure 3-4 The chromosome of offspring 1 generated by crossover in the embodiment of the present invention is

[0052] Figure 4-1 The crossover generation of offspring 2 in the embodiment of the present invention is represented by the parent chromosome

[0053] Figure 4-2 For the crossover generation of offspring 2 in the embodiment of the present invention - parent chromosome crossover point selection

[0054] Figure 4-3 Crossover to generate offspring 2 in the embodiment of the present invention - crossover of parent chromosome fragments

[0055] Figure 4-4 The chromosome of offspring 2 generated by crossover in the embodiment of the present invention is

[0056] Figure 5 The chromosomes of offspring 1 and offspring 2 generated after the crossover operation in the embodiment of the present invention are

[0057] Figure 6-1 The two-stage mutation operation in the embodiment of the present invention - parent mutation point selection

[0058] Figure 6-2 The two-stage mutation operation in the embodiment of the present invention - the mutation of the disassembly mode

[0059] Figure 6-3 This is a two-stage mutation operation in the embodiment of the present invention: finding the direct predecessor and direct successor of the task

[0060] Figure 6-4 This is a two-stage mutation operation in the embodiment of the present invention—inserting the task between the direct predecessor and the direct successor. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0062] On the contrary, the present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention as defined by the claims. Further, in order to make the public have a better understanding of the present invention, some specific details are described in detail in the detailed description of the present invention below. Those skilled in the art can fully understand the present invention without the description of these details.

[0063] The present invention provides a generative AI method to expand the disassembly data of products to be disassembled that account for a relatively small proportion. On the basis of known historical disassembly data, a generative AI method CTGAN for expanding tabular data is used to expand the disassembly data of products to be disassembled that account for a relatively small proportion of models, so as to better describe the disassembly situation of this type of products to be disassembled. The harmfulness and demand of the disassembly tasks of the products to be disassembled are finely classified, and a disassembly line balance model for mixed homogeneous products under a fine-grained mixed disassembly mode is established. On this basis, an improved multi-objective optimization algorithm based on NSGA-II is provided. According to the characteristics of the disassembly line balance problem, its encoding, decoding methods, initialization methods, and crossover, mutation and other operations are improved to make it suitable for solving the disassembly line balance problem.

[0064] The present invention provides an embodiment of an intelligent optimization disassembly line balancing method based on generative AI and an improved NSGA-II algorithm, such as Figure 1 As shown, the method includes:

[0065] (1) Collecting historical disassembly data and preprocessing the historical disassembly data;

[0066] (2) In the currently processed batch of orders, when the product type of the product to be disassembled accounts for less than 1 / 3 of the product type in the historical disassembly data, CTGAN is used to expand the disassembly data of the product to be disassembled to obtain the expanded disassembly data corresponding to the product to be disassembled that accounts for a relatively small proportion;

[0067] Among them, CTGAN (Conditional Tabular GAN) is a generative AI method specifically used for tabular data expansion. CTGAN takes the historical disassembly data preprocessed in step (1) as input, expands the data of the products to be disassembled that account for a relatively small proportion, and obtains the expanded disassembly data corresponding to the products to be disassembled that account for a relatively small proportion, so as to reduce the impact of the abnormal data of the product and better describe the disassembly of the product.

[0068] (3) Using the expanded disassembly data, fine-grained classification is performed on the harmfulness and demand of the disassembly tasks of the products to be disassembled, and a disassembly line balance solution model for mixed homogeneous products under a fine-grained mixed disassembly mode is established;

[0069] (4) Using the improved NSGA-II algorithm to solve the disassembly line balance solution model established in step (3) to obtain a solution to the current objective function; and by decoding the solution, a disassembly planning sequence is obtained.

[0070] In step (1) of this embodiment, the historical disassembly data includes order number, product type, product model, product quantity, disassembly time of different disassembly modes, disassembly energy consumption of different disassembly modes, disassembly cost of different disassembly modes, disassembly priority relationship constraint diagram of products, and disassembly feasible sequence of products;

[0071] Specifically, in this step, the collected historical disassembly data is the basis for establishing the model, and the historical disassembly data includes detailed information of completed disassembly orders.

[0072] In this embodiment, before using these data, preprocessing is required to ensure data quality and availability. The preprocessing includes the following steps: cleaning data, outlier processing, filling missing values, standardization and normalization, and feature engineering; the preprocessing process uses conventional techniques.

[0073] In step (3) of this embodiment, the disassembly line balance solution model for mixed homogeneous products under the fine-grained mixed disassembly mode is established, which includes: minimizing the number of workstations objective function , minimizing the workstation smoothing exponential objective function , minimize the cost objective function , minimize the energy consumption objective function , minimize the pollution index objective function .

[0074] The objective function formulas are:

[0075] ;

[0076] Where m is the workstation index, M is the total number of workstations, m=1,...,M; Indicates whether the mth workstation is turned on. When the workstation is turned on is equal to 1, when the workstation is shut down, is equal to 0.

[0077] ;

[0078] In the formula, i，l represents the disassembly task index, I represents the total number of disassembly tasks, , , ; j represents the position in the disassembly sequence, and J represents the total number of the largest disassembly tasks; Indicates the actual cycle time; Indicates whether task i is assigned to workstation m. When task i is assigned to workstation m is equal to 1 when task i is not assigned to workstation m is equal to 0; Representation Task l Whether to assign it to workstation m, when the task l When assigned to workstation m Equal to 1, when the task l When no workstation m is assigned is equal to 0; Indicates whether task i is in the jth position of workstation m. When task i is in the jth position of workstation m, Equal to 1, when task i is not in the jth position of workstation m =0; when task i is in the j+1th position of workstation m Equal to 1, when task i is not in the j+1th position of workstation m is equal to 0; Indicates whether there is a tool switch between task i and task l. When there is a tool switch between task i and task l Equal to 1, when there is no tool switching between task i and task l is equal to 0; The time required for a tool switch to occur;

[0079] represents the average disassembly time of task i, , K is the number of product types, k represents the product type, Indicates the disassembly ratio of k-type products; Indicates whether there is task i in product type k. If there is task i in product type k Equal to 1, when there is no task i in the k-type product is equal to 0; represents the disassembly time of product type k in disassembly task i; , , , represents the disassembly time of product type k in disassembly task i using non-destructive disassembly mode, represents the disassembly time of product type k in disassembly task i using destructive disassembly mode, Indicates the disassembly mode adopted by task i. When task i adopts the destructive mode Equal to 1, when task i adopts non-destructive mode is equal to 0.

[0080] ;

[0081] In the formula, Represents the cost per unit time of a workstation; It represents the additional cost per unit time for a workstation to process the primary hazardous product; It represents the additional cost per unit time for a workstation to process secondary harmful products; It represents the additional cost per unit time for a workstation to process the third-level harmful products; Indicates whether workstation m has a first-level dangerous task. If workstation m has a first-level dangerous task, then is 1, otherwise is 0; Indicates whether workstation m has a second-level dangerous task. If workstation m has a second-level dangerous task, then is 1, otherwise is 0; Indicates whether workstation m has a third-level dangerous task. If workstation m has a third-level dangerous task, then is 1, otherwise is 0;

[0082] represents the average disassembly cost of task i, , , represents the disassembly cost of product type k in disassembly task i, , , , is the disassembly cost of product type k in disassembly task i using non-destructive disassembly mode, is the disassembly cost of product type k in disassembly task i using destructive disassembly mode; Indicates the disassembly mode adopted by task i. When task i adopts the destructive mode Equal to 1, when task i adopts non-destructive mode is equal to 0.

[0083] ;

[0084] In the formula, It represents the energy consumption per unit time of a workstation; It represents the additional energy consumption per unit time of a workstation in processing primary harmful products; It represents the additional energy consumption per unit time of a workstation in processing secondary harmful products; It represents the additional energy consumption per unit time of a workstation in processing the third-level harmful products;

[0085] represents the average disassembly energy consumption of task i, , , represents the disassembly energy consumption of product type k in disassembly task i; , , , represents the disassembly energy consumption of product type k in disassembly task i using non-destructive disassembly mode, It represents the disassembly energy consumption of product type k in disassembly task i using destructive disassembly mode.

[0086] ;

[0087] In the formula, represents the average pollution index of task i; , , represents the pollution index of type k product in disassembly task i, , , , represents the contamination index of product type k in non-destructive disassembly mode in disassembly task i, It represents the pollution index of product type k in the destructive disassembly mode of disassembly task i.

[0088] In this embodiment, step (4) specifically includes the following steps:

[0089] (4.1) Coding: The coding method of NSGA-II is improved by using a matrix with 2 rows and 1 columns to encode the solution, such as Figure 2 As shown, the first row is the disassembly task number, and the second row is the task disassembly mode; the disassembly mode includes a destructive disassembly mode and a non-destructive disassembly mode, 0 represents a non-destructive mode, 1 represents a destructive mode, and the column number I is the total number of disassembly tasks;

[0090] (4.2) Initialization: In the initial solution generation, the tasks to be disassembled are arranged in ascending order according to the number of times they are selected. If the number of times they are selected is the same, they are arranged in descending order according to the hazard level, and the first task after the sorting is selected; the number of times selected refers to the number of times the task to be disassembled is selected in its position.

[0091] (4.3) Non-dominated sorting and crowding calculation: Perform non-dominated sorting on the current solution, calculate crowding, and select the individual to enter the next cycle;

[0092] (4.4) Crossover: Perform a crossover operation on the solution according to the characteristics of the problem. In the non-destructive NSGA-II method, the crossover operation is extremely random and cannot guarantee that the generated offspring will meet the constraints of this problem. The present invention improves the crossover operation of the non-destructive NSGA-II algorithm so that the generated offspring can meet the constraints. Using the two-point crossover method, two intersection points position1 and position2 are randomly generated, and the part between the intersection points position1 and position2 is the part to be crossed; the offspring chromosomes that meet the priority relationship are generated by exchanging the same gene sequences of the two parts to be crossed;

[0093] Specifically, in parent2, find the gene segment to be crossed between position1 and position2 in parent1. Arrange the gene segment to be crossed found in parent1 according to the gene order in parent2 to form a new gene segment, and use the new gene segment to replace the gene segment to be crossed in parent1 to form a new daughter chromosome child1. In the same way, re-arrange the gene segment to be crossed between position1 and position2 in parent2 according to its gene order in parent1 to generate a new daughter chromosome child2;

[0094] Since the crossover operation does not destroy the priority relationship between the processes, the new generation of chromosome sequences is also feasible. Figure 3-1 Two positions position1 and position2 are randomly selected from parent1 as intersection points, and the part between position1 and position2 is the part to be crossed; Figure 3-2 It is to match the genes in parent2 with the genes between the intersections in parent1; Figure 3-3 It rearranges the position of the part to be crossed in parent1 according to the position order of this part of genes in parent2; Figure 3-4 That is the result obtained after the crossover of parent1, which is child1. Figure 4-1 to Figure 4-4 The same idea is to perform a crossover operation on parent2. Figure 5 These are the offspring child1 and child2 obtained using the above crossover strategy.

[0095] (4.5) Mutation: Mutate the solution and adopt a two-stage mutation method to ensure that the generated offspring can meet the constraints. In the first stage, the position of the mutation point is randomly selected. If a task is harmful (specifically harmful to the environment or the health of workers) or has a recycling requirement, the disassembly mode of the corresponding task will not be changed. Otherwise, the disassembly mode of the corresponding task will be changed according to the randomly generated probability. In the second stage, the direct predecessor and direct successor of the task are found, and a position is selected between the direct predecessor and the direct successor, and the corresponding task is randomly inserted, such as Figure 6-1 to Figure 6-4 As shown; Figure 6-1 is the selection of mutation points, Figure 6-2 It is the mutation result of the disassembly mode after the first stage mutation. Figure 6-3 is to find the direct predecessor and direct successor of the task, and select a position between the direct predecessor and the direct successor for insertion. Figure 6-4 It is the offspring child obtained by adopting the two-stage mutation strategy.

[0096] (4.6) Loop: perform non-dominated sorting and crowding calculation on the current population, perform non-dominated sorting on the current solution, calculate crowding, and select individuals that can enter the next loop (selection criteria: individuals with low non-dominated levels after non-dominated sorting; individuals with high crowding at the same dominance level); until the termination condition is reached (specifically, in this embodiment, the termination condition is: reaching the maximum number of iterations 100), the loop ends, and the solution under the current objective function is obtained;

[0097] (4.7) Decoding: Arrange the tasks to be disassembled to the workstations in the order on the chromosome. If the current workstation is full, open the next workstation, and so on, until all the disassembly tasks are arranged.

[0098] The present invention adopts the improved intelligent optimization method NSGA-II algorithm to solve the disassembly line balance problem, which brings an efficient and flexible solution to the field of disassembly line balance; the improved intelligent optimization method NSGA-II algorithm can search the entire solution space and find a better disassembly sequence planning solution instead of just being limited to the local optimal solution. Using the improved intelligent optimization method NSGA-II algorithm to solve the disassembly line balance problem can allocate resources more reasonably and respond to different constraints more flexibly, thereby improving resource utilization efficiency and making the disassembly sequence planning more adapted to actual conditions.

[0099] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent optimization disassembly line balancing method based on CTGAN and improved NSGA-II algorithm, characterized in that: The method comprises: (1) Collect historical disassembly data; (2) In the currently processed batch of orders, when the product type of the product to be disassembled accounts for less than 1 / 3 of the product type in the historical disassembly data, CTGAN is used to expand the disassembly data of the product to be disassembled to obtain expanded disassembly data; (3) using the expanded disassembly data, fine-grained classification is performed on the harmfulness and demand of the disassembly tasks of the products to be disassembled, and a disassembly line balance solution model for mixed homogeneous products under a fine-grained mixed disassembly mode is established; (4) using an improved NSGA-II algorithm to solve the disassembly line balance solution model established in step (3) to obtain a solution to the current objective function; and obtaining a disassembly planning sequence by decoding the solution; Step (4) specifically includes the following steps: (4.1) Coding: The coding method of NSGA-II is improved, and a matrix of 2 rows and 1 columns is used to encode the solution. The first row is the task number to be disassembled, and the second row is the task disassembly mode. The disassembly mode includes destructive disassembly mode and non-destructive disassembly mode. 0 represents non-destructive mode, 1 represents destructive mode, and the number of columns I is the total number of disassembly tasks. (4.2) Initialization: In the initial solution generation, the tasks to be disassembled are arranged in ascending order according to the number of times they are selected. If the number of times they are selected is the same, they are arranged in descending order according to the hazard level; the first task after sorting is selected; (4.3) Non-dominated sorting and crowding calculation: Perform non-dominated sorting on the current solution, calculate crowding, sort in descending order according to crowding, and select the individuals to enter the next cycle; (4.4) Crossover: Use the two-point crossover method to randomly generate two crossover points, position1 and position2. The part between the crossover points position1 and position2 is the part to be crossed. Generate offspring chromosomes that meet the priority relationship by exchanging the same gene sequences of the two parts to be crossed. (4.5) Mutation: Mutate the solution and adopt a two-stage mutation method to ensure that the generated offspring can meet the constraints. In the first stage, the position of the mutation point is randomly selected. If a task is harmful or has a recycling requirement, the disassembly mode of the corresponding task will not be changed. Otherwise, whether to change the disassembly mode of the corresponding task is determined based on the randomly generated probability. In the second stage, the direct predecessor and direct successor of the task are found, and a position is selected between the direct predecessor and the direct successor to randomly insert the corresponding task. (4.6) Loop: Perform non-dominated sorting and crowding calculation on the current population, perform non-dominated sorting on the current solution, calculate crowding, and select individuals that can enter the next loop; until the termination condition is reached, the loop ends and the solution under the current objective function and constraint conditions is obtained; (4.7) Decoding: Arrange the tasks to be disassembled to the workstations in the order on the chromosome. If the current workstation is full, open the next workstation, and so on, until all the disassembly tasks are arranged.

2. The intelligent optimization disassembly line balancing method based on CTGAN and improved NSGA-II algorithm according to claim 1 is characterized in that: In step (1), the historical disassembly data includes order number, product type, product model, product quantity, disassembly time of different disassembly modes, disassembly energy consumption of different disassembly modes, disassembly cost of different disassembly modes, product disassembly priority relationship constraint diagram, and product disassembly feasible sequence.

3. The intelligent optimization disassembly line balancing method based on CTGAN and improved NSGA-II algorithm according to claim 1 is characterized in that: Step (1) also includes preprocessing the historical disassembly data; the preprocessing includes the steps of: cleaning data, outlier processing, filling missing values, standardization and normalization, and feature engineering.

4. The intelligent optimization disassembly line balancing method based on CTGAN and improved NSGA-II algorithm according to claim 1 is characterized in that: In step (3), the disassembly line balance solution model for mixed homogeneous products under the fine-grained mixed disassembly mode is established, including: minimizing the number of workstations objective function minf1, minimizing the workstation smoothness index objective function minf2, minimizing the cost objective function minf3, minimizing the energy consumption objective function minf4, and minimizing the pollution index objective function minf5.

5. The intelligent optimization disassembly line balancing method based on CTGAN and improved NSGA-II algorithm according to claim 4 is characterized in that: The formula for minimizing the number of workstations objective function minf1 is: Where m is the workstation index, M is the total number of workstations, m = 1, ..., M; u m Indicates whether the mth workstation is turned on. When the workstation is turned on, u m Equal to 1, when the workstation is shut down, u m is equal to 0.

6. The intelligent optimization disassembly line balancing method based on CTGAN and improved NSGA-II algorithm according to claim 5 is characterized in that: The formula for minimizing the workstation smoothing exponential objective function minf2 is: Where i, l represent the disassembly task index, I represents the total number of disassembly tasks, i=1,…,I,l=1,…,I,i≠l; j represents the position in the disassembly sequence, J represents the total number of the largest disassembly tasks; T P Indicates the actual cycle time; x im Indicates whether task i is assigned to workstation m. When task i is assigned to workstation m, x im is equal to 1, when task i is not assigned to workstation m im =0; x lm Indicates whether task l is assigned to workstation m. When task l is assigned to workstation m, x lm is equal to 1, when task l is not assigned to workstation m lm =0; y ijm Indicates whether task i is in the jth position of workstation m. When task i is in the jth position of workstation m, y ijm Equal to 1, when task i is not in the jth position of workstation m ijm is equal to 0; When task i is in the j+1th position of workstation m, y i(j+1)m Equal to 1, when task i is not in the j+1th position of workstation m i(j+1)m =0; z il Indicates whether there is a tool switch between task i and task l. When there is a tool switch between task i and task l, z il is equal to 1, when there is no tool switching between task i and task l il is equal to 0; t tool The time required for a tool switch to occur; t i represents the average disassembly time of task i, K is the number of product types, k represents the product type, ω k Indicates the disassembly ratio of type k products; δ ik Indicates whether there is task i in product type k. When there is task i in product type k, δ ik When δ is equal to 1, there is no task i in the k-type product ik is equal to 0; t ik represents the disassembly time of product type k in disassembly task i; in, represents the disassembly time of product type k in disassembly task i using non-destructive disassembly mode, represents the disassembly time of product type k in disassembly task i using destructive disassembly mode; r i Indicates the disassembly mode adopted by task i. When task i adopts the destructive mode, r i is equal to 1, when task i adopts non-destructive mode r i is equal to 0.

7. The intelligent optimization disassembly line balancing method based on CTGAN and improved NSGA-II algorithm according to claim 6 is characterized in that: The formula for minimizing the cost objective function minf3 is: In the formula, c α represents the cost per unit time of a workstation; c β1 represents the additional cost per unit time of a workstation for processing primary hazardous products; c β2 represents the additional cost per unit time of a workstation for processing secondary harmful products; c β3 Z represents the additional cost per unit time for a workstation to process the third-level harmful products; m1 Indicates whether workstation m has a first-level dangerous task. If workstation m has a first-level dangerous task, then Z m1 is 1, otherwise Z m1 is 0; Z m2 Indicates whether workstation m has a second-level dangerous task. If workstation m has a second-level dangerous task, then Z m2 is 1, otherwise Z m2 is 0; Z m3 Indicates whether workstation m has a third-level dangerous task. If workstation m has a third-level dangerous task, then Z m3 is 1, otherwise Z m3 is 0; represents the average disassembly cost of task i, c ik represents the disassembly cost of product type k in disassembly task i, is the disassembly cost of product type k in disassembly task i using non-destructive disassembly mode, is the disassembly cost of product type k in disassembly task i using destructive disassembly mode; r i Indicates the disassembly mode adopted by task i. When task i adopts the destructive mode, r i is equal to 1, when task i adopts non-destructive mode r i is equal to 0.

8. The intelligent optimization disassembly line balancing method based on CTGAN and improved NSGA-II algorithm according to claim 7 is characterized in that: The formula for minimizing the energy consumption objective function minf4 is: In the formula, e α Represents the energy consumption per unit time of a workstation; e β1 It represents the additional energy consumption per unit time of a workstation in processing primary harmful products; e β2 It represents the additional energy consumption per unit time of a workstation in processing secondary harmful products; e β3 It represents the additional energy consumption per unit time of a workstation in processing the third-level harmful products; e i represents the average disassembly energy consumption of task i, e ik represents the disassembly energy consumption of product type k in disassembly task i; represents the disassembly energy consumption of product type k in disassembly task i using non-destructive disassembly mode, It represents the disassembly energy consumption of product type k in disassembly task i using destructive disassembly mode.

9. The intelligent optimization disassembly line balancing method based on CTGAN and improved NSGA-II algorithm according to claim 8 is characterized in that: The formula for minimizing the pollution index objective function minf5 is: In the formula, represents the average pollution index of task i; p ik represents the pollution index of type k product in disassembly task i, represents the contamination index of product type k in non-destructive disassembly mode in disassembly task i, It represents the pollution index of product type k in the destructive disassembly mode of disassembly task i.

Images