Mixed flow man-machine cooperation disassembly line design method considering cost and energy consumption optimization

Abstract

The invention relates to the technical field of disassembly line balance, in particular to a mixed flow man-machine cooperation disassembly line design method considering cost and energy consumption optimization, which comprises the following steps of: 1, proposing a mixed flow man-machine cooperation disassembly line balance problem; 2, constructing a mixed integer programming model with the goal of minimizing the number of workstations, the idle time balance index, the disassembly cost and the disassembly energy consumption; and 3, designing an improved artificial fish swarm algorithm based on chemotactic operation for solving, introducing the chemotactic operation into foraging behaviors by the algorithm, providing an adaptive view strategy, and introducing a reverse learning mechanism in a population updating stage to generate a reverse solution and increase a feasible solution area. The fabric has excellent performance.

Description

Technical Field

[0001] The present invention relates to the technical field of disassembly line balance, and in particular to a mixed-flow human-machine collaborative disassembly line design method considering cost and energy consumption optimization. Background Art

[0002] With the vigorous development of social economy and the rapid progress of science and technology, various electronic and electrical products are constantly updated, resulting in a large number of waste products being eliminated. On the one hand, if the accumulation of these waste products is not handled in time, electronic waste will be generated, which contains harmful elements such as heavy metals. Direct disposal or incineration will cause pollution to the human body, the environment and the soil. On the other hand, there are some valuable and recyclable parts in waste products. By establishing a complete waste resource utilization system to recycle these resources, it will help promote the recycling of resources and achieve the goal of sustainable development of enterprises.

[0003] The disassembly of waste products is a key central link in the establishment and connection of forward logistics and reverse logistics of products in the circular economy. Efficient disassembly production methods can improve the efficiency of enterprises in recycling waste products. Disassembly lines have the advantages of high production efficiency, low cost, easy scale and automated disassembly. They are gradually replacing unitized disassembly and are being promoted and applied in disassembly companies. In the past, disassembly companies usually used manual operations to handle the disassembly tasks of waste products. Under this method, the labor intensity of workers was high and the disassembly efficiency was low, making it difficult to meet the needs of large-scale and diversified processing of waste products. In addition, some heavy and dangerous disassembly operations may also pose a potential threat to the physical and mental health of disassembly workers.

[0004] With the advancement of science and technology, the research and application of disassembly robots and automated disassembly systems have improved disassembly efficiency, reduced disassembly costs, and reduced workers' labor intensity and work risks. Although the progress of artificial intelligence and machine learning has improved the cognitive and learning capabilities of robots, robots still have limitations in disassembly operations with a high degree of uncertainty. In certain specific disassembly scenarios and in the face of complex disassembly tasks, robots still cannot completely replace manual disassembly operations. The human-robot collaboration (HRC) disassembly method combines human problem-solving ability, flexibility and adaptability with the load-bearing capacity, accuracy and durability of robots, thereby unleashing greater production potential.

[0005] Many scholars are currently conducting research on the application of human-robot collaborative work in the field of disassembly. For example, considering the non-disassembly of parts, some scholars proposed a human-robot collaborative disassembly line balancing problem (HRC-DLBP) with mixed local destructive operations. First, a mixed integer programming model was established for this problem to accurately obtain the number of workstations, smoothness index, cost, and the negative impact of noise pollution on workers. Then, an improved gray wolf optimization algorithm was proposed. Different cases were also used to verify that the proposed method is superior to three classic Swarm Intelligence methods and eleven other algorithms. Finally, the practicality of the proposed method was verified through an engine case. Some scholars have studied the balancing problem of a human-machine collaborative disassembly line under a combination of U-shaped and single straight-line layouts. Since it is necessary to consider the workload balancing problem of different disassembly lines, a mixed integer linear programming model is established for this problem with the goal of maximizing disassembly benefits. Different scale examples are introduced and CPLEX software is used to accurately solve them respectively. The comparative experimental results show that when disassembling products, the hybrid disassembly line of human-machine collaboration has obvious advantages over the disassembly line composed of only workers or only robots. Some scholars have proposed a model for balancing human-machine collaborative disassembly lines in view of the fact that existing human-machine collaborative disassembly research ignores time-varying human conditions such as safety, cognitive behavior, workload, and human posture changes. The model takes into account the load factors related to human participation, and aims to minimize the cycle time of the disassembly line and its related costs, while reducing the comprehensive smoothness index. At the same time, an improved multi-objective fruit fly optimization algorithm is used to solve and analyze an example of Mercedes battery pack disassembly, proving the efficiency and robustness of the improved multi-objective fruit fly optimization algorithm.

[0006] However, in the literature research on HRC-DLBP, most studies give priority to the optimization of the number of workstations, smoothness index, and disassembly time, and fewer studies consider the optimization of disassembly cost and disassembly energy consumption. For human-robot collaborative disassembly, some literature only considers robots replacing workers to perform some harmful and complex disassembly operations, and does not consider the disassembly operation characteristics and corresponding task operator allocation required for parts in waste products that need to be reused and remanufactured. In addition, waste products are often processed in large quantities in disassembly and recycling, and there is little research on the balance of human-robot collaborative disassembly lines under mixed flow conditions. Summary of the invention

[0007] The present invention provides a design method for a mixed-flow human-machine collaborative disassembly line taking into account cost and energy consumption optimization, which can overcome certain or some defects of the prior art.

[0008] According to the present invention, a mixed flow human-machine collaborative disassembly line design method considering cost and energy consumption optimization includes the following steps:

[0009] First, we propose MMHRC-DLBP for the mixed-flow human-machine collaborative disassembly line balancing problem;

[0010] Second, a mixed integer programming model is constructed to minimize the number of workstations, idle time balance index, disassembly cost and disassembly energy consumption;

[0011] 3. Design an improved artificial fish school algorithm based on chemotaxis operation. The algorithm introduces chemotaxis operation into foraging behavior and proposes an adaptive field of view strategy. In the population update stage, a reverse learning mechanism is introduced to generate reverse solutions to increase the feasible solution area.

[0012] As a preferred solution, the mixed flow human-machine collaborative disassembly line balancing problem MMHRC-DLBP is specifically:

[0013] All disassembly tasks are divided into three categories: disassembly tasks that can only be performed by robots, disassembly tasks that can only be performed by workers, and disassembly tasks that can be performed by both robots and workers. It is stipulated that disassembly tasks corresponding to hazardous parts are assigned to robots, disassembly tasks corresponding to parts that need to be reused or remanufactured are assigned to workers, and disassembly tasks corresponding to other parts are randomly assigned to any operator;

[0014] On the basis of classifying the various disassembly tasks of waste products and assigning task operators, combined with the derivation of the operation priority relationship constraints of the disassembly tasks of each waste product, the disassembly tasks of waste products are assigned to the various workstations of the human-machine collaborative disassembly line. In the process of workstation allocation of disassembly tasks, the constraint condition that the disassembly time of each workstation of the human-machine collaborative disassembly line cannot exceed the production rhythm of the human-machine collaborative disassembly line must be met. At the same time, the number of workstations opened on the human-machine collaborative disassembly line is minimized, the idle time balance index of each operator is minimized, the disassembly cost is minimized, and the disassembly energy consumption is minimized.

[0015] As a preferred method, the following assumptions are made when solving MMHRC-DLBP:

[0016] a) Consider the complete disassembly mode, that is, all parts of the scrapped product are removed, and the disassembly task corresponding to each part can only be assigned to one task operator to perform independently, that is, there is no task involving human-computer interaction;

[0017] b) The time required for workers and robots to disassemble each component is known and determined;

[0018] c) The walking time of the robot and the worker and the time required for tool replacement during the task execution are all included in the operation time of the corresponding disassembly task;

[0019] d) The physical implementation details of the disassembly process ensure effective collaboration between the two task operators, human and robot;

[0020] e) Ignore all uncertainties and emergencies during the disassembly process.

[0021] As a preference, the objective function of the mixed integer programming model is:

[0022] A) The first optimization goal is to minimize the number of workstations f1;

[0023]

[0024] B) The second optimization goal is to minimize the idle time balance index f2 of each operator; the task operation time assigned to each operator should be as close to the production rhythm as possible and as balanced as possible;

[0025]

[0026] C) The third optimization goal is the disassembly cost f3; the disassembly cost factors include four aspects: workstation startup cost, robot use cost, worker disassembly cost and robot disassembly cost; the disassembly cost calculation formula is as follows:

[0027]

[0028] D) The fourth optimization goal is the disassembly energy consumption f4; the energy consumption required during the disassembly operation is divided into robot working energy consumption, robot standby energy consumption and workstation operation fixed energy consumption; the disassembly energy consumption calculation formula is as follows:

[0029]

[0030] The constraints of the mixed integer programming model are:

[0031]

[0032] Where v is the type of waste product; n is the total number of disassembly tasks for all waste products; k is the workstation opening number, k∈{1,2,…,K}, K=n; w is the type of disassembly operator, if the operator is a worker, then w=1, if the operator is a robot, then w=2; i,j are task numbers, i,j∈{1,2,…,n}; The time required for disassembly of task i assigned to operator w; s i h is the time when disassembly task i starts operation; i is a 0-1 variable. If the disassembly task i has a hazardous attribute, then h i =1, otherwise h i =0;di is a 0-1 variable. If disassembly task i has the attribute of reuse or remanufacturing demand, then d i =1, otherwise d i =0;x ikw is a 0-1 variable. If disassembly task i is assigned to operator w in workstation k, then x ikw =1, otherwise x ikw =0;S k It is a 0-1 variable. If the workstation is turned on, S k =1, otherwise S k =0; is a 0-1 variable. If operator w is assigned to workstation k, then otherwise Q ijkw is a 0-1 variable. If tasks i and j are assigned to the same operator at the same workstation and task i is disassembled before task j, then Q ijkw =1, otherwise Q ijkw =0;P ij is a 0-1 variable. If task i is the immediate predecessor of task j, then P ij =1, otherwise P ij =0, according to the disassembly relationship between tasks, the disassembly priority relationship matrix TP = (P ij ) n×n ; M is a large number; C k is the fixed cost of opening a workstation; C r is the fixed cost of using the robot; C rt C is the cost of robot disassembly operation per unit time; pt The cost of disassembly work per worker unit time; i is the energy consumption of the robot per unit disassembly operation time; e m is the standby energy consumption per unit time of the robot; e r The energy consumption required for the robot to perform a reversing operation on the product; f Other energy consumption, including workstation fixed energy consumption, lighting, and ventilation energy consumption;

[0033] Formula (5) represents the task allocation constraint; Formula (6) and Formula (7) represent the workstation opening constraint and continuous opening constraint respectively; Formula (8) represents the job priority relationship constraint between the disassembly tasks of the same product; Formula (9) represents the constraint on the number of operators in the workstation. Due to the limited area of ​​the workstation, in order to prevent interference between operators in the same workstation, it is limited that there are at most two operators working at the same time in the same workstation, that is, one worker and one robot; Formula (10) specifies the starting disassembly operation time of task i; Formula (11) specifies the disassembly completion time of task i. The completion time should be less than the end time of the workstation; Formula (12) and Formula (13) stipulate that when any two tasks are assigned to the same operator, the task operator cannot perform disassembly tasks in parallel; Formula (14) stipulates that if any two disassembly tasks i and j are assigned to the same operator and task j is executed before task i, task i can only be disassembled after task j is disassembled; Formula (15) means that when two tasks with a close priority relationship are assigned to the same workstation, even if the two tasks are assigned to different operators, task j can only be disassembled after task i is disassembled.

[0034] As a preferred embodiment, the specific implementation steps of the improved artificial fish swarm algorithm based on chemotaxis operation are as follows:

[0035] 1) Initialize parameters and calculate the objective function;

[0036] 2) Screen non-inferior solutions, update the external archive set Q, and determine the optimal solution for each single objective;

[0037] 3) Set the number of iterations gen = 1 and start iteration;

[0038] 4) Select an artificial fish X cur ;

[0039] 5) Update the average field of view of the school of fish;

[0040] 6) If X cur If clustering occurs, proceed to the next step. If X cur If a rear-end collision occurs, proceed to step 14);

[0041] 7) Determine whether the grouping is successful. If so, proceed to the next step. If not, foraging behavior occurs;

[0042] 8) Select non-inferior solutions and update Q;

[0043] 9) Reverse learning generates reverse solutions;

[0044] 10) Screen non-inferior solutions and update the optimal solution for each single objective;

[0045] 11) Determine whether the current artificial fish number reaches the population size fish_num, if so, proceed to the next step, if not, return to step 4);

[0046] 12) Update population;

[0047] 13) Determine if gen < max_gen, where max_gen is the maximum number of iterations. If so, output the optimal solution. If not, gen = gen + 1, and return to step 4);

[0048] 14) Determine whether the rear-end collision is successful. If so, proceed to step 8). If not, foraging behavior occurs.

[0049] Preferably, in the foraging behavior, the steps are as follows:

[0050] 1.1) Set the number of times t = 1, artificial fish X cur Perform cross-operation and update external file Q fora ;

[0051] 1.2) Determine whether t>n try , if not, then t=t+1, return to the crossover operation, if so, proceed to the next step;

[0052] 1.3) Determine Q fora＝ X cur If it is equal, perform chemotaxis operation first, and then proceed to step 8); if it is not equal, proceed directly to step 8).

[0053] Preferably, the specific steps of chemotaxis operation are as follows:

[0054] 2.1) Set the number of targets m = 1, the maximum number of rotations n_step = 1, and select the first single-target optimal solution X m ;

[0055] 2.2) Rotation operation: X cur With X m Perform fragment exchange to generate a new solution X new ;

[0056] 2.3) Determine X new Is it better than X? cur , if so, then X new =X cur , proceed to the next step, if not, randomly generate a solution and X cur Perform a rotation;

[0057] 2.4)n_step=n_step+1;

[0058] 2.5) Determine whether n_step>n_step_max. If not, return to step 2.2. If yes, proceed to the next step.

[0059] 2.6) m = m + 1;

[0060] 2.7) m>m_max, if not, return to step 2.2), if yes, proceed to step 8).

[0061] As a preferred method, the problem coding includes two parts. The first part is the disassembly task sequence of the product. A one-dimensional array of 1 to n is used to correspond to the disassembly tasks. The coding of the task sequence must meet the constraints of the job priority relationship between the disassembly tasks. The second part represents different types of disassembly operators. An array of the same length as the first part is randomly generated. The array contains only two elements, 0 and 1, to represent workers and robots respectively. Then, according to the attribute constraints of each disassembly task, the assigned disassembly operators are adjusted to ensure that the task attributes match the disassembly operators one by one, that is, parts with reuse and demand attributes must be assigned to workers for disassembly, and parts with hazards must be assigned to robots for disassembly.

[0062] The decoding process is to assign the tasks in the disassembly sequence to the task operators in the workstations. The specific steps are as follows:

[0063] Step 1: Input the disassembly task sequence X, the comprehensive priority relationship matrix TP, and the disassembly information matrix KB;

[0064] Step 2: Open the first workstation WS=1, record the current task sequence position p=1, record the remaining allocatable time RT of the workstation, and set RT=CT, where CT is the takt time;

[0065] Step 3: Determine the type of disassembly operator w corresponding to task i assigned to position p in the sequence. After determining the corresponding disassembly operator, extract the disassembly time corresponding to task i at position p according to the disassembly information matrix KB.

[0066] Step 4: Determine all the immediate predecessors of task i and find the end time f of the last task j that is assigned to be disassembled among all immediate predecessors j ;

[0067] Step 5: Determine whether task j is in the current workstation WS. If so, go to Step 6. If not, determine the end time f of the previous task y of operator w corresponding to task i in the current workstation. y , if the remaining time that the current workstation can allocate to task i Then the start time of task i is s i =f y Otherwise, a new workstation WS=WS+1 is opened, and the start time of task i is si = CT*(WS-1);

[0068] Step 6: According to the end time of the previous task j and the end time of the previous task y in the current workstation, determine the remaining time that can be allocated to task i in the current workstation RT = min (CT*WS-f j ,CT*WS-f y ),like Then a new workstation WS=WS+1 is opened, and the start time of task i is s i = CT*(WS-1), otherwise the start time of task i is s i =max(f j ,f y );

[0069] Step 7: If p ≥ n, all tasks are assigned, otherwise set p = p + 1 and return to Step 2.

[0070] Preferably, in the adaptive field of view strategy, a large field of view Visual is set at the beginning of the algorithm iteration so that the artificial fish has a larger search space when making behavior selections; before each iteration, the current artificial fish calculates the distance between itself and other fish within the field of view and calculates the average value, which is used as the field of view for the next iteration.

[0071] As a preferred method, in the reverse learning mechanism, is a feasible solution at the tth iteration, represents the j-th dimension component of the solution sequence, where is a random value between [0,1], as shown in the following formula:

[0072]

[0073] for The reverse solution of express The component of the j-th dimension on the solution sequence;

[0074] According to the characteristics of the DLBP problem, the disassembly task sequence is taken as a feasible solution, and the disassembly task number is mapped to the corresponding dimensional component; the maximum and minimum values ​​of each dimensional component are defined as the positions of its adjacent preceding and succeeding tasks respectively; when determining the reverse population, each task is inserted into its reverse component position.

[0075] Based on the application potential of human-robot collaboration in disassembly production lines, and taking into account the diversity of waste products to be disassembled, the present invention proposes a mixed-model human-robot collaborative disassembly line. Aiming at the mixed-model human-robot collaborative disassembly line balancing problem (Mixed-Model Human-Robot Collaborative Disassembly Line Balancing Problem, MMHRC-DLBP), a multi-objective MIP mathematical model of the proposed problem is constructed with the optimization objectives of minimizing the number of workstations, idle time balance index, disassembly cost and disassembly energy consumption. Considering the solution characteristics of the mixed-model human-robot collaborative disassembly line balancing problem, an improved artificial fish swarming algorithm based on chemotaxis operation (Chemotaxis operation-basedArtificial Fish Swarming Algorithm, CAFSA) is designed. The proposed algorithm introduces chemotaxis operation in foraging behavior, and proposes an adaptive field of view strategy to improve the global optimization performance of the algorithm. In the population update stage, a reverse learning mechanism is introduced to generate reverse solutions to increase the feasible solution area and accelerate the convergence of the algorithm. The present invention has excellent performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 It is a flow chart of a method for designing a mixed-flow human-machine collaborative disassembly line considering cost and energy consumption optimization in an embodiment;

[0077] Figure 2 Schematic diagram of a mixed flow human-machine collaborative disassembly line in an embodiment;

[0078] Figure 3 A schematic diagram for generating a comprehensive disassembly operation priority relationship matrix in an embodiment;

[0079] Figure 4 It is a flowchart of improving the artificial fish swarm algorithm in the embodiment;

[0080] Figure 5 This is a schematic diagram of the coding of the human-machine collaborative disassembly task sequence in the embodiment;

[0081] Figure 6 Schematic diagram of reverse solution in different spatial dimensions in the embodiment. DETAILED DESCRIPTION

[0082] In order to further understand the content of the present invention, the present invention is described in detail in conjunction with the accompanying drawings and embodiments. It should be understood that the embodiments are only for explaining the present invention and are not intended to limit it.

[0083] Example

[0084] like Figure 1As shown, this embodiment provides a mixed flow human-machine collaborative disassembly line design method considering cost and energy consumption optimization, which includes the following steps:

[0085] First, we propose MMHRC-DLBP for the mixed-flow human-machine collaborative disassembly line balancing problem;

[0086] Second, a mixed integer programming model is constructed to minimize the number of workstations, idle time balance index, disassembly cost and disassembly energy consumption;

[0087] 3. Design an improved artificial fish school algorithm based on chemotaxis operation. The algorithm introduces chemotaxis operation into foraging behavior and proposes an adaptive field of view strategy. In the population update stage, a reverse learning mechanism is introduced to generate reverse solutions to increase the feasible solution area.

[0088] Mixed-flow human-robot collaborative disassembly line balancing problem (MMHRC-DLBP)

[0089] Specifically:

[0090] The subject who performs the disassembly tasks in the traditional manual disassembly line and the robot disassembly line is single, and all disassembly tasks are performed by the same operator, and there is no classification of disassembly tasks. However, the disassembly tasks in the human-machine collaborative disassembly line are performed by workers and robots together. Due to the differences in the conditions for completing different disassembly tasks, the disassembly tasks cannot be directly assigned to workers or robots, and the same disassembly task assigned to different task operation objects will affect the balance of the human-machine collaborative disassembly line. For this reason, this embodiment divides all disassembly tasks into three categories: disassembly tasks that are only allowed to be performed by robots, disassembly tasks that are only allowed to be performed by workers, and disassembly tasks that can be performed by both robots and workers. It is also stipulated that the disassembly tasks corresponding to hazardous parts are assigned to robots, the disassembly tasks corresponding to parts that need to be reused or remanufactured are assigned to workers, and the disassembly tasks corresponding to the remaining parts are randomly assigned to any operator.

[0091] like Figure 2 The figure shows a schematic diagram of a mixed-flow human-machine collaborative disassembly line, in which two types of task operators, workers and robots, simultaneously perform disassembly of mixed-flow products in each workstation. The human-machine collaborative disassembly line balance problem studied in this embodiment is to classify the various disassembly tasks of waste products and assign task operators, and combine the derivation of the job priority relationship constraints of the various disassembly tasks of each waste product to assign the various disassembly tasks of the waste products to the various workstations of the human-machine collaborative disassembly line. In the process of station allocation of the disassembly task, the constraint condition that the disassembly time of each workstation of the human-machine collaborative disassembly line cannot exceed the production rhythm of the human-machine collaborative disassembly line must be met, and at the same time, the number of workstations opened in the human-machine collaborative disassembly line is minimized, the idle time balance index of each operator is minimized, the disassembly cost is minimized, and the disassembly energy consumption is minimized.

[0092] For mixed flow disassembly, this paper divides each product to be disassembled into the smallest disassembly unit according to the quantity ratio of 1:1. To ensure the correctness of the disassembly sequence, the disassembly operation priority relationship matrix of each product is combined into a comprehensive disassembly operation priority relationship matrix through direct sum operation. A waste product with 5 disassembly tasks and a waste product with 6 disassembly tasks are used as examples to illustrate the construction of the comprehensive disassembly operation priority relationship matrix. The disassembly operation priority relationship diagrams of the two products are shown as follows: Figure 3 As shown in the figure, the blue circles are the disassembly tasks corresponding to the parts with reuse / remanufacturing demand attributes, and the green circles are the disassembly tasks corresponding to the parts with hazardous attributes. The disassembly priority matrices of the two products are A1 and A2 respectively. By performing a direct sum operation on A1 and A2, the comprehensive disassembly priority relationship matrix of the two products is obtained.

[0093] Assumptions:

[0094] a) Consider the complete disassembly mode, that is, all parts of the scrapped product are removed, and the disassembly task corresponding to each part can only be assigned to one task operator to perform independently, that is, there is no task involving human-computer interaction;

[0095] b) The time required for workers and robots to disassemble each component is known and determined;

[0096] c) The walking time of the robot and the worker and the time required for tool replacement during the task execution are all included in the operation time of the corresponding disassembly task;

[0097] d) The physical implementation details of the disassembly process ensure effective collaboration between the two task operators, human and robot;

[0098] e) Ignore all uncertainties and emergencies during the disassembly process.

[0099] Mixed Integer Programming Model

[0100] The objective function is:

[0101] A) The first optimization goal is to minimize the number of workstations f1;

[0102]

[0103] B) The second optimization goal is to minimize the idle time balance index f2 of each operator; the task operation time assigned to each operator should be as close to the production rhythm as possible and as balanced as possible;

[0104]

[0105] C) The third optimization goal is the disassembly cost f3; the disassembly cost factors include four aspects: workstation startup cost, robot use cost, worker disassembly cost and robot disassembly cost; the disassembly cost calculation formula is as follows:

[0106]

[0107] D) The fourth optimization goal is the disassembly energy consumption f4; the energy consumption required during the disassembly operation is divided into robot working energy consumption, robot standby energy consumption and workstation operation fixed energy consumption; the disassembly energy consumption calculation formula is as follows:

[0108]

[0109] The constraints are:

[0110]

[0111] Where v is the type of waste product; n is the total number of disassembly tasks for all waste products; k is the workstation opening number, k∈{1,2,…,K}, K=n; w is the type of disassembly operator, if the operator is a worker, then w=1, if the operator is a robot, then w=2; i,j are task numbers, i,j∈{1,2,…,n}; The time required for disassembly of task i assigned to operator w; s i h is the time when disassembly task i starts operation; i is a 0-1 variable. If the disassembly task i has a hazardous attribute, then h i =1, otherwise h i =0;d i is a 0-1 variable. If disassembly task i has the attribute of reuse or remanufacturing demand, then d i =1, otherwise d i =0;x ikw is a 0-1 variable. If disassembly task i is assigned to operator w in workstation k, then x ikw =1, otherwise x ikw =0;S k It is a 0-1 variable. If the workstation is turned on, S k =1, otherwise S k =0; is a 0-1 variable. If operator w is assigned to workstation k, then otherwise Q ijkw is a 0-1 variable. If tasks i and j are assigned to the same operator at the same workstation and task i is disassembled before task j, then Q ijkw =1, otherwise Q ijkw =0;P ijis a 0-1 variable. If task i is the immediate predecessor of task j, then P ij =1, otherwise P ij =0, according to the disassembly relationship between tasks, the disassembly priority relationship matrix TP = (P ij ) n×n ; M is a large number; C k is the fixed cost of opening a workstation; C r is the fixed cost of using the robot; C rt C is the cost of robot disassembly operation per unit time; pt The cost of disassembly work per worker unit time; i is the energy consumption of the robot per unit disassembly operation time; e m is the robot's standby energy consumption per unit time; e r The energy consumption required for the robot to perform a reversing operation on the product; f Other energy consumption, including workstation fixed energy consumption, lighting, and ventilation energy consumption;

[0112] Formula (5) represents the task allocation constraint; Formula (6) and Formula (7) represent the workstation opening constraint and continuous opening constraint respectively; Formula (8) represents the job priority relationship constraint between the disassembly tasks of the same product; Formula (9) represents the constraint on the number of operators in the workstation. Due to the limited area of ​​the workstation, in order to prevent interference between operators in the same workstation, it is limited that there are at most two operators working at the same time in the same workstation, that is, one worker and one robot; Formula (10) specifies the starting disassembly operation time of task i; Formula (11) specifies the disassembly completion time of task i. The completion time should be less than the end time of the workstation; Formula (12) and Formula (13) stipulate that when any two tasks are assigned to the same operator, the task operator cannot perform disassembly tasks in parallel; Formula (14) stipulates that if any two disassembly tasks i and j are assigned to the same operator and task j is executed before task i, task i can only be disassembled after task j is disassembled; Formula (15) means that when two tasks with a close priority relationship are assigned to the same workstation, even if the two tasks are assigned to different operators, task j can only be disassembled after task i is disassembled.

[0113] Improved artificial fish swarm algorithm based on chemotaxis operation

[0114] The AFSA algorithm is a new type of intelligent optimization algorithm. Its basic idea is to simulate some ecological behaviors of real fish schools in the water, such as foraging, clustering and chasing, so that the algorithm can find a better solution during the exploration process. Considering the solution characteristics of the mixed flow human-machine collaborative disassembly line balance problem studied in this embodiment, the AFSA algorithm is improved by introducing chemotaxis operation in the foraging behavior, and an adaptive field of view strategy is proposed to improve the global optimization performance of the algorithm. In the population update stage, a reverse learning mechanism is introduced to generate reverse solutions to increase the feasible solution area and accelerate the convergence of the algorithm. The overall process of the algorithm is as follows: Figure 4 The specific implementation steps are as follows:

[0115] 1) Initialize parameters and calculate the objective function;

[0116] 2) Screen non-inferior solutions, update the external archive set Q, and determine the optimal solution for each single objective;

[0117] 3) Set the number of iterations gen = 1 and start iteration;

[0118] 4) Select an artificial fish X cur ;

[0119] 5) Update the average field of view of the school of fish;

[0120] 6) If X cur If clustering occurs, proceed to the next step. If X cur If a rear-end collision occurs, proceed to step 14);

[0121] 7) Determine whether the grouping is successful. If so, proceed to the next step. If not, foraging behavior occurs;

[0122] 8) Select non-inferior solutions and update Q;

[0123] 9) Reverse learning generates reverse solutions;

[0124] 10) Screen non-inferior solutions and update the optimal solution for each single objective;

[0125] 11) Determine whether the current artificial fish number reaches the population size fish_num, if so, proceed to the next step, if not, return to step 4);

[0126] 12) Update population;

[0127] 13) Determine if gen < max_gen, where max_gen is the maximum number of iterations. If so, output the optimal solution. If not, gen = gen + 1, and return to step 4);

[0128] 14) Determine whether the rear-end collision is successful. If so, proceed to step 8). If not, foraging behavior occurs.

[0129] In foraging behavior, the steps are as follows:

[0130] 1.1) Set the number of times t = 1, artificial fish X cur Perform cross-operation and update external file Q fora ;

[0131] 1.2) Determine whether t>n try , if not, then t=t+1, return to the crossover operation, if so, proceed to the next step;

[0132] 1.3) Determine Q fora＝ X cur If it is equal, perform chemotaxis operation first, and then proceed to step 8); if it is not equal, proceed directly to step 8).

[0133] Chemotaxis

[0134] This paper introduces the chemotaxis operation in the bacterial foraging algorithm into the foraging behavior of the artificial fish swarm algorithm, and uses the strong local optimization ability of the chemotaxis operator to jump out of the local optimum and find the global optimal solution. Chemotaxis refers to the movement of bacteria by random selection, and then judging whether it is moving in that direction by comparing the fitness function before and after the movement.

[0135] The chemotaxis operation description of bacterium i per unit step is shown in formula (16):

[0136]

[0137] Among them, θ i (j,k,l) ​​represents the position of the i-th bacterium at the j-th chemotaxis operation, the k-th replication operation and the l-th migration operation, C(i) is the swimming step length of the chemotaxis operation, and φ(j) is the random unit length direction vector of the chemotaxis operation. Δ(i) is a random number between [-1,1]. θ i (j+1,k,l) ​​represents the position of the i-th bacterium after its j+1-th swimming.

[0138] When the foraging behavior of the artificial fish swarm algorithm fails, the current solution cur To make the chemotaxis move in a better direction, set the optimal solution set for each single target, select the optimal solution for each single target, exchange the fragments and continuously update the solution until the maximum number of swimming times is reached. The specific steps are as follows:

[0139] 2.1) Set the number of targets m = 1, the maximum number of rotations n_step = 1, and select the first single-target optimal solution X m ;

[0140] 2.2) Rotation operation: X cur With Xm Perform fragment exchange to generate a new solution X new ;

[0141] 2.3) Determine X new Is it better than X? cur , if so, then X new =X cur , proceed to the next step, if not, randomly generate a solution and X cur Perform a rotation;

[0142] 2.4)n_step=n_step+1;

[0143] 2.5) Determine whether n_step>n_step_max. If not, return to step 2.2. If yes, proceed to the next step.

[0144] 2.6) m = m + 1;

[0145] 2.7) m>m_max, if not, return to step 2.2), if yes, proceed to step 8).

[0146] Encoding and Decoding

[0147] The problem coding consists of two parts. The first part is the product disassembly task sequence, which uses a one-dimensional array of 1 to n to correspond to the disassembly tasks. The coding of the task sequence must meet the job priority relationship constraints between the disassembly tasks. The second part represents different types of disassembly operators. First, an array of the same length as the first part is randomly generated. The array contains only two elements, 0 and 1, to represent workers and robots respectively. Then, according to the attribute constraints of each disassembly task, the assigned disassembly operators are adjusted to ensure that the task attributes match the disassembly operators one by one, that is, parts with reuse and demand attributes must be assigned to workers for disassembly, and parts with hazards must be assigned to robots for disassembly. Figure 5 A schematic diagram of the coding sequence for a human-machine collaborative disassembly task.

[0148] The decoding process is to assign the tasks in the disassembly sequence to the task operators in the workstations. The specific steps are as follows:

[0149] Step 1: Input the disassembly task sequence X, the comprehensive priority relationship matrix TP, and the disassembly information matrix KB;

[0150] Step 2: Open the first workstation WS=1, record the current task sequence position p=1, record the remaining allocatable time RT of the workstation, and set RT=CT, where CT is the takt time;

[0151] Step 3: Determine the type of disassembly operator w corresponding to task i assigned to position p in the sequence. After determining the corresponding disassembly operator, extract the disassembly time corresponding to task i at position p according to the disassembly information matrix KB.

[0152] Step 4: Determine all the immediate predecessors of task i and find the end time f of the last task j that is assigned to be disassembled among all immediate predecessors j ;

[0153] Step 5: Determine whether task j is in the current workstation WS. If so, go to Step 6. If not, determine the end time f of the previous task y of operator w corresponding to task i in the current workstation. y , if the remaining time that the current workstation can allocate to task i Then the start time of task i is s i =f y Otherwise, a new workstation WS=WS+1 is opened, and the start time of task i is s i = CT*(WS-1);

[0154] Step 6: According to the end time of the immediately preceding task j and the end time of the previous task y in the current workstation, determine the remaining time that can be allocated to task i in the current workstation RT = min (CT*WS-f j ,CT*WS-f y ),like Then a new workstation WS=WS+1 is opened, and the start time of task i is s i = CT*(WS-1), otherwise the start time of task i is s i =max(f j ,f y );

[0155] Step 7: If p ≥ n, all tasks are assigned, otherwise set p = p + 1 and return to Step 2.

[0156] Adaptive Vision Strategy

[0157] In the artificial fish swarm algorithm, choosing a suitable field of view of the artificial fish is crucial to the convergence performance of the algorithm. In general, if the field of view is small, the tail-chasing and grouping behaviors will be weaker, the search ability in the vicinity will be improved, and the foraging behavior will be more outstanding. On the contrary, if the field of view is larger, the search space becomes wider, because the artificial fish will also conduct a lot of foraging and random activities in the large field of view, which will slow down the convergence of the algorithm. Therefore, in order to overcome the defects of the artificial fish swarm algorithm and based on the characteristics of the research problem, an improved method of adaptive field of view is proposed. The specific approach includes: at the beginning of the algorithm iteration, a large field of view Visual is set so that the artificial fish has a larger search space when making behavior selections; before each iteration, the current artificial fish calculates the distance between other fish and itself within the field of view and calculates the average value, which is used as the field of view for the next iteration. Assume that there are artificial fish fish1, fish2, fish3, fish4, and fish5 in the domain S. The distances between fish1 and the other four fish are d1, d2, d3, and d4. At this time, the field of view of fish1 is (d1+d2+d3+d4) / 4.

[0158] At the beginning of the algorithm, a larger field of view is set so that the algorithm has a larger search space in the early stage of iteration, thereby increasing the probability of the algorithm finding the global optimal value. As the algorithm continues to iterate, its field of view is shrinking. Therefore, the algorithm searches a local area of ​​the current artificial fish in the later stage. At this time, the foraging behavior of the artificial fish algorithm is active. In order to avoid the algorithm from falling into the local optimal state due to the narrowing of the field of view in the later stage, the chemotaxis operation in the bacterial algorithm is introduced in the foraging behavior to increase the ability to jump out of the local optimal state, thereby finding the global optimal solution.

[0159] Reverse learning mechanism

[0160] The quality of the solution is usually poor in the process of randomly generating the initial population. The quality of the solution of the newly generated population in the population update stage has a great impact on the convergence of the algorithm. In order to retain the better solution in the algorithm iteration process, the CAFSA algorithm is screened based on the Pareto idea and the crowding distance mechanism to obtain all solutions and the non-inferior solutions obtained by screening are retained in the external archive set Q. Then the reverse learning mechanism is introduced to generate reverse solutions for all updated individuals obtained by the algorithm after updating the external archive set Q. The reverse learning mechanism is a mechanism that can effectively expand the search space and cover the feasible solution area. It has been successfully used in intelligent algorithms. This mechanism generates a reverse solution based on the existing solution, and then compares the target fitness values ​​of the two. If the fitness value of the reverse solution is better, then the reverse solution will be selected for iterative optimization in the next iteration, and vice versa.

[0161] like Figure 6The figure shows an intuitive schematic diagram of feasible solutions and their reverse solutions in different spatial dimensions. As can be seen from the figure above, the reverse point of any point in different spatial dimensions is unique. In the multi-objective optimization algorithm, as the number of iterations increases, the variables in one dimension are constantly changing. In order to continuously generate reverse solutions during the iteration process, a dynamic reverse optimization strategy is proposed. is a feasible solution at the tth iteration, represents the j-th dimension component of the solution sequence, where is a random value between [0,1], as shown in the following formula:

[0162]

[0163] for The reverse solution of express The component of the j-th dimension on the solution sequence;

[0164] In combination with the characteristics of the DLBP problem, the disassembly task sequence is used as a feasible solution, and the disassembly task number is mapped to the corresponding dimensional component; the maximum and minimum values ​​of each dimensional component are defined as the positions of its adjacent preceding and succeeding tasks respectively; when determining the reverse population, each task is inserted into its reverse component position. The following is a pseudo code design for generating a reverse solution:

[0165]

[0166] Model Validation

[0167] In order to verify the performance of the CAFSA algorithm, two experiments were conducted. In the first part, the correctness of the constructed MIP model and the feasibility of the CAFSA algorithm were verified by examples. In the second part, the performance of the CAFSA algorithm on mixed-flow human-machine collaborative disassembly examples of different scales was experimentally analyzed. First, a mixed-flow human-machine collaborative disassembly example consisting of two products (P7 and P5) containing 7 tasks and 5 tasks respectively was introduced to verify the correctness of the constructed MIP model. The CAFSA algorithm was used to solve the example and the solution results were compared with the exact solution obtained by the LINGO software to verify the feasibility of the algorithm. The operating environment of each software is: Windows 10 system, Inter (R) Core (TM) i5-9400 CPU @ 2.90GHz, 16GB memory. The parameters of the CAFSA algorithm for this small-scale example are set as follows: fish_num = 30, max_gen = 30, try_num = 10, δ = 0.8 and n_step = 8, where the initial field of view is set to visual = 15 and the cycle time is CT = 40s. The relevant disassembly data information of the two products to be disassembled, P7 and P5, are shown in Table 1.

[0168] Table 2 lists the optimal values ​​of each objective obtained by LINGO for this example and the Pareto non-inferior solution set containing 8 non-inferior solutions obtained by the CAFSA algorithm for this small-scale human-machine collaborative disassembly example. It can be obtained from Table 2 that for objective f1, 6 non-inferior solutions in the solution set obtain the minimum value of 2 on this objective; for f2, the 8th non-inferior solution obtained by the CAFSA algorithm obtains the minimum value of 49 on this objective; for f3, the 5th non-inferior solution obtained by the CAFSA algorithm obtains the minimum value of 17.52 on this objective; and for f4, the 6th non-inferior solution obtained by the CAFSA algorithm obtains the minimum value of 796 on this objective. In addition, from the comparison of solution time, the time taken by LINGO to solve the four objectives of this example is 127s, 7534s, 2960s, and 39s respectively, while the time taken by the CAFSA algorithm to obtain this set of Pareto non-inferior solutions is only 30.19s. A comprehensive comparative analysis shows that compared with LINGO software, which can only obtain the exact solution of one target at a time and takes a long time to solve, the CAFSA algorithm can obtain a solution set including the exact solution of multiple targets at one time in a short time. In summary, the correctness of the constructed MIP model and the feasibility and efficiency of the CAFSA algorithm are verified through a small-scale example.

[0169] Table 1 Relevant disassembly data information of P7 and P5

[0170]

[0171] Table 2 Comparison of solution results between LINGO and CAFSA algorithms

[0172]

[0173]

[0174] Case Analysis

[0175] In order to further verify the performance of the proposed CAFSA algorithm in solving mixed-flow human-machine collaborative disassembly cases of different scales, 9 products from existing literature, namely Model1(P5), Model2(P7), Model3(P8), Model4(P10), Model5(P11), Model6(P22), Model7(P24), Model8(P25), and Model9(P27), are introduced. These 9 products are randomly combined to form a mixed-flow human-machine collaborative disassembly case set of different scales containing 30 different product combinations. According to the total number of disassembly tasks, the case set is divided into 10 small-scale cases (disassembly tasks are less than or equal to 30), 10 medium-scale cases (disassembly tasks are greater than 30 and less than or equal to 70), and 10 large-scale cases (disassembly tasks are greater than 70). The proposed CAFSA algorithm is used to solve these 30 cases respectively, and the three algorithms of AFSA, NSGAII, and TLBO are introduced for comparative analysis. The parameters of the CAFSA algorithm under different scales are set as follows: fish_num=30, max_gen=30, try_num=10 and n_step=8 (small scale); fish_num=50, max_gen=50, try_num=15 and n_step=8 (medium scale); fish_num=80, max_gen=100, try_num=15 and n_step=8 (large scale). The comparison algorithms use the set parameters in their respective references.

[0176] The above four algorithms solve the 30 groups of examples 10 times respectively, and introduce four multi-objective evaluation indicators, NR, C, GD and IGD, to comprehensively evaluate the quality of the Pareto solutions obtained by each algorithm. Table 3-5 shows the evaluation index values ​​of NR, C, GD and IGD corresponding to the Pareto non-inferior solution sets obtained by each algorithm for 10 solutions to each example of small, medium and large scales. The first row of each algorithm under each evaluation index represents the average value of the corresponding evaluation index obtained by the algorithm for 10 solutions to a certain example, and the second row represents the minimum value of the corresponding evaluation index obtained by the algorithm for 10 solutions to a certain example. When calculating the GD and IGD indicators, the true Pareto optimal solution is obtained by further screening all the Pareto non-inferior solutions obtained by the four algorithms for solving the example.

[0177] The results show that the average values ​​of the four evaluation indicators obtained by the CAFSA algorithm are better than those of the other three compared algorithms overall, and the effect of the algorithm becomes more significant as the task scale gradually increases.

[0178] From the perspective of NR index, in the 10 small-scale examples, the average value of the results obtained by the CAFSA algorithm is the best among the four algorithms. Although the search range of each algorithm in the small-scale examples is small and the performance difference is relatively insignificant, after 10 independent tests, the average value of the NR index of the results of the small-scale examples obtained by CAFSA is better than that of the other three algorithms. In the 10 examples under the medium-scale task, except for the 1st and 2nd groups, the average value of the NR index of the CAFSA algorithm is the best in the remaining 8 groups. In the 10 large-scale examples, the NSGAII algorithm has the best NR index result in the first group, and the CAFSA algorithm has the best NR index results in the remaining 9 groups. Through the analysis of the results of the small, medium and large scales, it can be concluded that for the NR index, the NR index of the results obtained by CAFSA is obviously superior among the four algorithms.

[0179] From the perspective of the C index, in the small-scale examples, CAFSA has the largest average value in groups 6, 7, 8, and 10, while NSGAII has the largest average value of the C index in groups 2 and 4, and TLBO is the algorithm in groups 1, 3, and 5. In the medium-scale examples, the optimal value of the C index in group 1 is the TLBO algorithm, groups 3 and 5 are NSGAII algorithms, and the rest are CAFSA. For the maximum value of the C index in the 10 medium-scale examples, CAFSA has 6 groups of results that are 1, NSGAII and TLBO have 2 groups of results that are 1, and AFSA has no results. In the large-scale examples, except for the first group where the optimal value of the C index is obtained by NSGAII, the remaining 9 groups of optimal values ​​are all CAFSA, and the maximum value of the C index in 10 groups of examples is 1 for CAFSA, 6 for NSGAII, 1 for AFSA, and no for TLBO. Among the three scales, CAFSA, which has the largest mean value of the C index, accounts for the largest proportion, and CAFSA also has the largest number of maximum values ​​among the 30 groups of examples, followed by NSGAII, TLBO and AFSA.

[0180] From the perspective of GD index, in the small-scale examples, CAFSA has 4 groups with the best performance, namely 2, 4, 7, and 8, while NSGAII has 3 groups with the best performance, namely 5, 9, and 10. The best average of the remaining 3 groups of GD index is TLBO. In the medium-scale examples, NSGAII has the best results in groups 1, 6, and 8, while TLBO has the best results only in group 2. The best values ​​of the remaining 6 groups of examples are all obtained by CAFSA. However, in the large-scale 10 groups of examples, the average of GD index obtained by CAFSA is the best.

[0181] Finally, by comparing the IGD indicators of the solutions of the four algorithms, the average and optimal values ​​of the IGD indicators of the solutions obtained by the CAFSA algorithm for different scale cases are dominant, indicating that the solution obtained by CAFSA has excellent convergence and distribution.

[0182] By comparing the values ​​of the four evaluation indicators, it can be found that for small-scale examples, CAFSA does not have a significant advantage in each evaluation indicator. This is due to the limitation of the number of tasks, which leads to a small search range for the algorithm and relatively insignificant performance differences among the algorithms. As the scale increases, in the 20 groups of medium and large-scale examples, it can be found that the dominance ratio of the four evaluation indicator values ​​corresponding to the solution results of the CAFSA algorithm is increasingly better than that of the other three algorithms, which means that the CAFSA algorithm can generate a Pareto solution with higher quality and converge more to the ideal Pareto frontier.

[0183] Table 3 Calculation results of four indicators for small-scale examples

[0184]

[0185]

[0186] Table 4 Calculation results of the four indicators of the scale example

[0187]

[0188]

[0189] Table 5 Calculation results of four indicators for large-scale examples

[0190]

[0191]

[0192] In summary, for the solution results of large, medium and small scales, each evaluation indicator of the CAFSA algorithm is outstanding. Therefore, it can be concluded that from the perspectives of solution diversity, distribution and algorithm convergence, CAFSA has a great advantage in solving the mixed-flow human-machine collaborative disassembly line balance problem.

[0193] This embodiment aims at the disadvantages of traditional manual disassembly methods that are gradually exposed in the disassembly production of large-scale and diversified waste products. The human-machine collaborative disassembly method is introduced into the disassembly line, and a mixed flow human-machine collaborative disassembly line balance problem considering cost and energy consumption optimization is proposed. A multi-objective MIP model with the optimization objectives of minimizing the number of workstations, idle time balance index, disassembly cost and disassembly energy consumption is constructed, and a CAFSA algorithm is proposed for solution. In order to reduce the influence of the fish school field of view parameters on the convergence of the algorithm, the algorithm proposes an adaptive field of view strategy for improvement; when the artificial fish school foraging behavior fails, the algorithm is searched through the chemotaxis operator operation to avoid falling into the local optimum; the solution space is increased by generating a reverse solution through the reverse learning operation of the solution of the external archive set, and the convergence of the algorithm is accelerated. By using the proposed CAFSA algorithm to solve the small-scale mixed flow human-machine collaborative disassembly example, and comparing the solution results with the results obtained by the LINGO software, the correctness of the constructed MIP model and the effectiveness of the proposed CAFSA algorithm are verified. Furthermore, the proposed CAFSA algorithm is used to solve 30 sets of mixed-flow human-machine collaborative disassembly examples of different scales, and the TLBO, AFSA, and NSGAII algorithms are introduced to compare the solution results of the four algorithms for 30 sets of examples of different scales with the proposed CAFSA algorithm on the four multi-objective evaluation indicators of NR, C, GD, and IGD. The comparison results prove the excellent solution performance of the proposed CAFSA.

[0194] The present invention and its embodiments are described schematically above, and the description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. Therefore, if a person skilled in the art is inspired by it and designs a structural method and an embodiment similar to the technical solution without creativity without departing from the purpose of the invention, they shall all fall within the protection scope of the present invention.

Claims

1. A design method for a mixed-flow human-machine collaborative disassembly line considering cost and energy consumption optimization, characterized in that: The following steps are involved: First, a mixed-flow human-machine collaborative disassembly line balancing problem MMHRC-DLBP is proposed; Second, a mixed integer programming model is constructed to minimize the number of workstations, idle time balance index, disassembly cost and disassembly energy consumption; 3. Design an improved artificial fish school algorithm based on chemotaxis operation. The algorithm introduces chemotaxis operation into foraging behavior and proposes an adaptive field of view strategy. In the population update stage, a reverse learning mechanism is introduced to generate reverse solutions to increase the feasible solution area.

2. According to claim 1, a design method for a mixed flow human-machine collaborative disassembly line considering cost and energy consumption optimization is characterized in that: The mixed flow human-machine collaborative disassembly line balancing problem MMHRC-DLBP is specifically: All disassembly tasks are divided into three categories: disassembly tasks that can only be performed by robots, disassembly tasks that can only be performed by workers, and disassembly tasks that can be performed by both robots and workers. It is stipulated that disassembly tasks corresponding to hazardous parts are assigned to robots, disassembly tasks corresponding to parts that need to be reused or remanufactured are assigned to workers, and disassembly tasks corresponding to other parts are randomly assigned to any operator; On the basis of classifying the various disassembly tasks of waste products and assigning task operators, combined with the derivation of the operation priority relationship constraints of the disassembly tasks of each waste product, the disassembly tasks of waste products are assigned to the various workstations of the human-machine collaborative disassembly line. In the process of workstation allocation of disassembly tasks, the constraint condition that the disassembly time of each workstation of the human-machine collaborative disassembly line cannot exceed the production rhythm of the human-machine collaborative disassembly line must be met. At the same time, the number of workstations opened on the human-machine collaborative disassembly line is minimized, the idle time balance index of each operator is minimized, the disassembly cost is minimized, and the disassembly energy consumption is minimized.

3. The method for designing a mixed-flow human-machine collaborative disassembly line considering cost and energy consumption optimization according to claim 2, characterized in that: When solving MMHRC-DLBP, the following assumptions are made: a) Consider the complete disassembly mode, that is, all parts of the scrapped product are removed, and the disassembly task corresponding to each part can only be assigned to one task operator to perform independently, that is, there is no task involving human-computer interaction; b) The time required for workers and robots to disassemble each component is known and determined; c) The walking time of the robot and the worker and the time required for tool replacement during the task execution are all included in the operation time of the corresponding disassembly task; d) The physical implementation details of the disassembly process ensure effective collaboration between the two task operators, human and robot; e) Ignore all uncertainties and emergencies during the disassembly process.

4. The method for designing a mixed-flow human-machine collaborative disassembly line considering cost and energy consumption optimization according to claim 3, characterized in that: The objective function of the mixed integer programming model is: A) The first optimization goal is to minimize the number of workstations f1; B) The second optimization goal is to minimize the idle time balance index f2 of each operator; the task operation time assigned to each operator should be as close to the production rhythm as possible and as balanced as possible; C) The third optimization goal is the disassembly cost f3; the disassembly cost factors include four aspects: workstation startup cost, robot use cost, worker disassembly cost and robot disassembly cost; the disassembly cost calculation formula is as follows: D) The fourth optimization goal is the disassembly energy consumption f4; the energy consumption required during the disassembly operation is divided into robot working energy consumption, robot standby energy consumption and workstation operation fixed energy consumption; The calculation formula for disassembly energy consumption is as follows: The constraints of the mixed integer programming model are: Where v is the type of waste product; n is the total number of disassembly tasks for all waste products; k is the workstation opening number, k∈{1,2,…,K}, K=n; w is the type of disassembly operator, if the operator is a worker, then w=1, if the operator is a robot, then w=2; i,j are task numbers, i,j∈{1,2,…,n}; The time required for disassembly of task i assigned to operator w; s i h is the time when disassembly task i starts operation; i is a 0-1 variable. If the disassembly task i has a hazardous attribute, then h i =1, otherwise h i =0;d i is a 0-1 variable. If disassembly task i has the attribute of reuse or remanufacturing demand, then d i =1, otherwise d i =0;x ikw is a 0-1 variable. If disassembly task i is assigned to operator w in workstation k, then x ikw =1, otherwise x ikw =0;S k It is a 0-1 variable. If the workstation is turned on, S k =1, otherwise S k =0; is a 0-1 variable. If operator w is assigned to workstation k, then otherwise Q ijkw is a 0-1 variable. If tasks i and j are assigned to the same operator at the same workstation and task i is disassembled before task j, then Q ijkw =1, otherwise Q ijkw =0;P ij is a 0-1 variable. If task i is the immediate predecessor of task j, then P ij =1, otherwise P ij =0, according to the disassembly relationship between tasks, the disassembly priority relationship matrix TP = (P ij ) n×n ; M is a large number; C k is the fixed cost of opening a workstation; C r is the fixed cost of using the robot; C rt C is the cost of robot disassembly operation per unit time; pt The cost of disassembly work per worker unit time; i is the energy consumption of the robot per unit disassembly operation time; e m is the standby energy consumption per unit time of the robot; e r The energy consumption required for the robot to perform a reversing operation on the product; f Other energy consumption, including workstation fixed energy consumption, lighting, and ventilation energy consumption; Formula (5) represents the task allocation constraint; Formula (6) and Formula (7) represent the workstation opening constraint and continuous opening constraint respectively; Formula (8) represents the job priority relationship constraint between the disassembly tasks of the same product; Formula (9) represents the constraint on the number of operators in the workstation. Due to the limited area of ​​the workstation, in order to prevent interference between operators in the same workstation, it is limited that there are at most two operators working at the same time in the same workstation, that is, one worker and one robot; Formula (10) specifies the starting disassembly operation time of task i; Formula (11) specifies the disassembly completion time of task i. The completion time should be less than the end time of the workstation; Formula (12) and Formula (13) stipulate that when any two tasks are assigned to the same operator, the task operator cannot perform disassembly tasks in parallel; Formula (14) stipulates that if any two disassembly tasks i and j are assigned to the same operator and task j is executed before task i, task i can only be disassembled after task j is disassembled; Formula (15) means that when two tasks with a close priority relationship are assigned to the same workstation, even if the two tasks are assigned to different operators, task j can only be disassembled after task i is disassembled.

5. The method for designing a mixed-flow human-machine collaborative disassembly line considering cost and energy consumption optimization according to claim 4, characterized in that: The specific implementation steps of the improved artificial fish swarm algorithm based on chemotaxis operation are as follows: 1) Initialize parameters and calculate the objective function; 2) Screen non-inferior solutions, update the external archive set Q, and determine the optimal solution for each single objective; 3) Set the number of iterations gen = 1 and start iteration; 4) Select an artificial fish X cur ; 5) Update the average field of view of the school of fish; 6) If X cur If clustering occurs, proceed to the next step. If X cur If a rear-end collision occurs, proceed to step 14); 7) Determine whether the grouping is successful. If so, proceed to the next step. If not, foraging behavior occurs; 8) Select non-inferior solutions and update Q; 9) Reverse learning generates reverse solutions; 10) Screen non-inferior solutions and update the optimal solution for each single objective; 11) Determine whether the current artificial fish number reaches the population size fish_num, if so, proceed to the next step, if not, return to step 4); 12) Update population; 13) Determine if gen < max_gen, where max_gen is the maximum number of iterations. If so, output the optimal solution. If not, gen = gen + 1, and return to step 4); 14) Determine whether the rear-end collision is successful. If so, proceed to step 8). If not, foraging behavior occurs.

6. The method for designing a mixed-flow human-machine collaborative disassembly line considering cost and energy consumption optimization according to claim 5, characterized in that: In foraging behavior, the steps are as follows: 1.1) Set the number of times t = 1, artificial fish X cur Perform cross-operation and update external file Q fora ; 1.2) Determine whether t>n try , if not, then t=t+1, return to the crossover operation, if so, proceed to the next step; 1.3) Determine Q fora＝ X cur If it is equal, perform chemotaxis operation first, and then proceed to step 8); if it is not equal, proceed directly to step 8).

7. The method for designing a mixed-flow human-machine collaborative disassembly line considering cost and energy consumption optimization according to claim 6, characterized in that: The specific steps of chemotaxis operation are as follows: 2.1) Set the number of targets m = 1, the maximum number of rotations n_step = 1, and select the first single-target optimal solution X m ; 2.2) Rotation operation: X cur With X m Perform fragment exchange to generate a new solution X new ; 2.3) Determine X new Is it better than X? cur , if so, then X new =X cur , proceed to the next step, if not, randomly generate a solution and X cur Perform a rotation; 2.4)n_step=n_step+1; 2.5) Determine whether n_step>n_step_max. If not, return to step 2.

2. If yes, proceed to the next step. 2.6) m = m + 1; 2.7) m>m_max, if not, return to step 2.2), if yes, proceed to step 8).

8. The method for designing a mixed-flow human-machine collaborative disassembly line considering cost and energy consumption optimization according to claim 7, characterized in that: The problem coding consists of two parts. The first part is the product disassembly task sequence, which uses a one-dimensional array of 1 to n to correspond to the disassembly tasks. The coding of the task sequence must meet the job priority relationship constraints between the disassembly tasks. The second part represents different types of disassembly operators. First, an array of the same length as the first part is randomly generated. The array contains only two elements, 0 and 1, to represent workers and robots respectively. Then, according to the attribute constraints of each disassembly task, the assigned disassembly operators are adjusted to ensure that the task attributes match the disassembly operators one by one, that is, parts with reuse and demand attributes must be assigned to workers for disassembly, and parts with hazards must be assigned to robots for disassembly. The decoding process is to assign the tasks in the disassembly sequence to the task operators in the workstations. The specific steps are as follows: Step 1: Input the disassembly task sequence X, the comprehensive priority relationship matrix TP, and the disassembly information matrix KB; Step 2: Open the first workstation WS=1, record the current task sequence position p=1, record the remaining allocatable time RT of the workstation, and set RT=CT, where CT is the takt time; Step 3: Determine the type of disassembly operator w corresponding to task i assigned to position p in the sequence. After determining the corresponding disassembly operator, extract the disassembly time corresponding to task i at position p according to the disassembly information matrix KB. Step 4: Determine all the immediate predecessors of task i and find the end time f of the last task j that is assigned to be disassembled among all immediate predecessors j ; Step 5: Determine whether task j is in the current workstation WS. If so, go to Step 6. If not, determine the end time f of the previous task y of operator w corresponding to task i in the current workstation. y , if the remaining time that the current workstation can allocate to task i Then the start time of task i is s i =f y Otherwise, a new workstation WS=WS+1 is opened, and the start time of task i is s i = CT*(WS-1); Step 6: According to the end time of the immediately preceding task j and the end time of the previous task y in the current workstation, determine the remaining time that can be allocated to task i in the current workstation RT = min (CT*WS-f j ,CT*WS-f y ),like Then a new workstation WS=WS+1 is opened, and the start time of task i is s i = CT*(WS-1), otherwise the start time of task i is s i =max(f j ,f y ); Step 7: If p ≥ n, all tasks are assigned, otherwise set p = p + 1 and return to Step 2.

9. The method for designing a mixed-flow human-machine collaborative disassembly line considering cost and energy consumption optimization according to claim 8, characterized in that: In the adaptive vision strategy, at the beginning of the algorithm iteration, a large visual range Visual is set so that the artificial fish has a larger search space when making behavior selections; Before each iteration, the current artificial fish calculates the distance between itself and other fish within its field of view and takes the average value, which is used as the field of view for the next iteration.

10. The method for designing a mixed-flow human-machine collaborative disassembly line considering cost and energy consumption optimization according to claim 9, characterized in that: In the reverse learning mechanism, is a feasible solution at the tth iteration, represents the j-th dimension component of the solution sequence, where is a random value between [0,1], as shown in the following formula: for The reverse solution of express The component of the j-th dimension on the solution sequence; According to the characteristics of the DLBP problem, the disassembly task sequence is taken as a feasible solution, and the disassembly task number is mapped to the corresponding dimensional component; the maximum and minimum values ​​of each dimensional component are defined as the positions of its adjacent preceding and succeeding tasks respectively; when determining the reverse population, each task is inserted into its reverse component position.

Images