Multi-target man-machine collaborative disassembly line balance problem solving method

Abstract

The invention relates to the technical field of disassembly line balance, and provides a multi-target man-machine cooperation disassembly line balance problem solving method comprising the following steps: 1, providing a man-machine cooperation disassembly line balance problem HRC-DLBP considering part remanufacturing demand characteristics and hazard characteristics, and aiming at a task operator distribution problem, providing a multi-target man-machine cooperation disassembly line balance problem HRC-DLBP; the method comprises the following steps: establishing a mapping constraint mechanism between a disassembly task and a task operator based on remanufacturing requirements and hazard characteristics of all parts of a waste product; 2, constructing a mixed integer programming model aiming at the proposed problem, wherein the mixed integer programming model takes minimization of the number of workstations, minimization of task operator idle time balance indexes, minimization of demand indexes and risk indexes as optimization targets; and 3, solving through a multi-objective improved teaching optimization (MTLO) algorithm fused with an embedded disturbance strategy based on a random transformation task operator. According to the invention, the balance problem of the multi-target man-machine collaborative disassembly line can be well solved.

Description

Technical Field

[0001] The present invention relates to the technical field of disassembly line balancing, and in particular to a multi-objective human-machine collaborative disassembly line balancing problem solving method. Background Art

[0002] The rapid development of economy and technology has led to a continuous increase in the number of waste products. These waste products contain recyclable parts and materials, but at the same time, these waste products also contain substances that may cause damage and pollution to the human body and the environment. With the continuous strengthening of resource and environmental constraints, the development of a green, low-carbon circular economy has become a fundamental way to break resource and environmental constraints, respond to climate change, and cultivate new economic growth points. The shortage of resources and the degradation of the ecological environment are key constraints that hinder the sustainable development of contemporary society. Therefore, how to deal with these waste products in a timely and efficient manner is a top priority for governments of various countries. With the introduction and promulgation of relevant policies, regulations and rules and regulations, the recycling and reuse industry of waste products has also continued to develop, providing a platform for the standardized and standardized recycling and reuse of waste products. Disassembly is a key link in the recycling and reuse of waste products, and it is also the central link in the establishment and connection of product forward logistics and reverse logistics.

[0003] At present, the disassembly of waste products mainly relies on manpower. However, with the explosive growth of the number of waste products, the disassembly workload has become increasingly intensive, which has aggravated the fatigue of disassembly workers and caused a sharp decline in disassembly production efficiency. Therefore, it is difficult to cope with the increasingly heavy disassembly workload relying solely on manpower. In addition, some parts may contain harmful substances that may cause damage and pollution to human health or the ecological environment, as well as some disassembly operations that do not meet ergonomic requirements during the production process, which will also have an adverse effect on the physical and mental health of workers. The development of collaborative robots has promoted the rise and development of the human-robot collaboration (HRC) disassembly operation mode, making up for the shortcomings of the worker and robot disassembly mode. The HRC disassembly mode combines the advantages of human problem-solving ability, flexibility, and adaptability with the robot's load-bearing capacity, accuracy, and durability, and has greater production potential.

[0004] There are two types of task operators on the HRC disassembly line: workers and robots. These two types of task operators have their own task execution capabilities and applicability characteristics. For example, robots are more suitable for operations with heavy loads, high repetition frequencies, and some hazardous tasks; while workers are more suitable for more complex and delicate disassembly tasks, such as the disassembly of rusted and unremovable fasteners, deformed and aged parts, etc. When the disassembly tasks corresponding to the same parts are assigned to different task operators, their corresponding disassembly methods, disassembly time spent, and disassembly tools used may all be different. These differences will significantly affect the planning of the disassembly sequence and the allocation of disassembly task workstations, and thus affect the workload distribution among the task operators on the HRC disassembly line and the overall efficiency of the HRC disassembly line. Therefore, the Human-Robot Collaborative Disassembly Line Balancing Problem (HRC-DLBP) is a multi-level and multi-dimensional decision-making problem that integrates the coupling optimization of the three sub-problems of disassembly task operator allocation, disassembly sequence planning, and disassembly task workstation allocation, and has a high solution complexity.

[0005] In the existing literature on HRC-DLBP, some studies on task operator allocation simply randomly assign disassembly tasks to workers or robots, and most studies divide disassembly tasks into two categories: hazardous tasks and complex tasks. There is no literature on the task operator allocation problem for parts with remanufacturing demand characteristics. Regarding the optimization objectives, most studies prioritize the optimization of the number of workstations, smoothness index, disassembly cost, and disassembly time. Few studies consider the disassembly priority of hazardous parts and remanufactured parts. Summary of the invention

[0006] The present invention provides a multi-objective human-machine collaborative disassembly line balancing problem solution method, which can overcome certain or some defects of the prior art.

[0007] A multi-objective human-machine collaborative disassembly line balancing problem solving method according to the present invention comprises the following steps:

[0008] First, a human-robot collaborative disassembly line balancing problem (HRC-DLBP) is proposed, which takes into account the remanufacturing demand and hazard characteristics of parts. Aiming at the task operator allocation problem, a mapping constraint mechanism between disassembly tasks and task operators is established based on the remanufacturing demand and hazard characteristics of each part of waste products.

[0009] Second, a mixed integer programming model is constructed to optimize the problem by minimizing the number of workstations, minimizing the idle time balance index of task operators, and minimizing the demand index and the danger index.

[0010] 3. The solution is obtained through a multi-objective improved teaching optimization MTLO algorithm that integrates an embedded perturbation strategy based on randomly transformed task operators.

[0011] As a preferred embodiment, in step 1, a human-machine collaborative disassembly line balancing problem HRC-DLBP considering the demand characteristics and hazard characteristics of component remanufacturing is:

[0012] There are two types of task operators on the HRC disassembly line: workers and robots. Each workstation on the HRC disassembly line can only be equipped with one worker and one robot to perform disassembly tasks in parallel. In order to solve the operator allocation problem of disassembly tasks in HRC-DLBP, all disassembly tasks are divided into three categories according to the hazards and remanufacturing demand characteristics of the parts corresponding to the disassembly tasks:

[0013] a) Disassembly tasks corresponding to hazardous parts are only allowed to be performed by robots;

[0014] b) The disassembly tasks corresponding to the parts that need to be remanufactured are only allowed to be performed manually;

[0015] c) Other tasks that can be performed by robots or workers;

[0016] Taking into account that workers and robots have their own task execution capabilities and characteristics, a mapping constraint mechanism is constructed to characterize the matching relationship between task operators and disassembly tasks with different characteristic attributes; on this basis, combined with the derivation of the priority relationship constraints among the disassembly tasks of waste products, under the constraint of following the production rhythm, the disassembly tasks are assigned to the task operators at each workstation of the HRC disassembly line, and a balancing scheme for the HRC disassembly line is sought with the minimum number of disassembly workstations, the minimum idle time balance index of the task operators, the minimum danger index and the minimum demand index.

[0017] As a preferred method, the following assumptions are made when solving HRC-DLBP:

[0018] 1) Consider the complete disassembly mode, and the disassembly task corresponding to each component can only be assigned to one task operator to perform independently, that is, there is no task involving human-computer interaction;

[0019] 2) The time it takes for workers and robots to disassemble each component is known;

[0020] 3) The operator's walking time and tool changing time are included in the disassembly operation time;

[0021] 4) The physical implementation details of the disassembly process ensure effective collaboration between workers and robots;

[0022] 5) Ignore all uncertainties and emergency situations in the HRC disassembly production process.

[0023] Preferably, in step 2, the objective function of the mixed integer programming model is:

[0024]

[0025]

[0026] Formulas (1)-(4) describe four optimization objectives respectively: minimizing the number of HRC disassembly workstations, minimizing the idle time balance index of task operators, minimizing the demand index, and minimizing the hazard index;

[0027] The model constraints are:

[0028]

[0029]

[0030] Where i, j represent the disassembly task index; p represents the disassembly task sequence number index; k represents the HRC disassembly workstation index; w represents the task operator, if the task operator is a robot, w = 1, if the task operator is a worker, w = 2; CT represents the production line cycle; t i w represents the operation time required for task operator w to perform disassembly task i; D i represents the demand index. If the parts corresponding to disassembly task i are remanufacturable, then D i =1; otherwise, D i =0; H i represents the hazard index. If the component corresponding to the disassembly task i is hazardous, then H i =1; otherwise, H i =0; t bi represents the time when task i starts to execute; M represents a positive number; PD represents the disassembly priority relationship matrix PD = [pd ij ] I×I , if task i is followed by task j, pd ij =1; otherwise, pd ij =0; Z k Represents a binary variable. If HRC disassembly station k is turned on, Z k =1; otherwise, Z k =0; U k w represents a binary variable. If task operator w is assigned to workstation k, U k w =1; otherwise U kw =0;x ik w represents a binary variable, if task i is assigned to task operator w in workstation k, then x ik w =1; otherwise x ik w =0;Q ijk w represents a binary variable. If disassembly tasks i and j are assigned to the same task operator w in the same workstation k, and disassembly task i is executed before task j, then Q ijk w =1; otherwise, Q ijk w =0;s ip represents a binary variable. If the disassembly sequence number of disassembly task i is p, s ip =1; otherwise s ip =0;

[0031] Formula (5) indicates that any disassembly task must be assigned and can only be assigned to a task operator in a certain workstation; Formula (6) indicates that the disassembly task corresponding to hazardous parts must be assigned to robot disassembly; Formula (7) ensures that the disassembly task corresponding to parts with remanufacturing requirements must be assigned to workers for disassembly; Formula (8) indicates that if a workstation is open, the workstation must be equipped with at least one type of task operator, and can be equipped with at most two different types of task operators; Formulas (9)-(10) indicate that if a workstation is equipped with a certain type of task operator, there must be at least one disassembly task assigned to the task operator, and the number of tasks assigned to the task operator cannot exceed the total number of disassembly tasks; Formulas (11)-(12) indicate that if a human-machine collaborative disassembly workstation is open, it must be guaranteed that at least one disassembly task is assigned to the workstation, and the total number of tasks assigned to the workstation cannot exceed the total number of disassembly tasks; Formula (13) indicates that the workstations are opened in sequence and there are no empty workstations;

[0032] Formula (14) indicates that if disassembly task i is the immediate predecessor of task j, the start execution time of disassembly task j shall not be earlier than the execution completion time of disassembly task i; Formula (15) indicates that if any two different disassembly tasks i and j are assigned to the same task operator of a workstation, and task i is executed before task j, the start execution time of disassembly task j shall not be earlier than the completion time of task i; Formulas (16)-(17) indicate that if two disassembly tasks i and j are assigned to the same task operator in the same workstation, the two disassembly tasks i and j cannot be operated in parallel at the same time; Formula (18) indicates that if two different disassembly tasks i and j are assigned to the same task operator of the workstation, and task i is executed before task j, the start execution time of task j must be greater than the sum of the start time of the corresponding workstation and the disassembly time of task i; Formula (19) indicates that any task i must be disassembled within the cycle time; Formula (20) stipulates that the start time of any disassembly task i is a positive real number.

[0033] Preferably, in step 3, the specific implementation steps of the MTLO algorithm are as follows:

[0034] Step 1: Initialize the algorithm parameters, including the number of individuals in the population N, the number of iterations Iter, the number of disturbances rd, the learning phase 1 parameter Num_ls1 and the learning phase 2 parameter Lk; input the problem parameters: the operation time of each disassembly task, the operation priority relationship between disassembly tasks, and the scheduled HRC disassembly line production rhythm;

[0035] Step 2: Randomly generate the initial population, calculate the objective function value of each individual in the population, and screen the Pareto non-dominated individuals on this basis;

[0036] Step 3: Set the iteration count to 1, that is, iter = 1;

[0037] Step 4: During the teaching phase, for each student individual, a random neighborhood search operation of a two-point crossover operation is performed to update the current student individual;

[0038] Step 5: Implement the rd non-repetitive perturbation strategy for each student individual obtained in step 4 to update the current student individual;

[0039] Step 6: In learning phase 1, perform a local search operation based on a single-point insertion operation driven by a greedy mechanism on each student individual obtained in step 5 to generate Num_ls1 neighborhood individuals, and perform a Pareto comparison between the current student individual and the newly generated neighborhood individuals; use the selected Pareto non-dominated individuals to update the current student individual; then implement the rd non-repeated perturbation strategy for each student individual to update the current student individual;

[0040] Step 7: In the learning phase 2, a local search strategy of two-point exchange operation is performed on each student individual obtained in step 6, and the knowledge level of these student individuals is continuously updated; on this basis, a perturbation strategy of rd times without repetition is implemented for each student individual to update the current student individual;

[0041] Step 8: Update the current Pareto non-dominated individuals by performing Pareto comparison on all current student individuals;

[0042] Step 9: iter=iter+1, if iter<=Iter, return to step 4; otherwise, go to step 10;

[0043] Step 10: Screen and obtain the final Pareto non-dominated individuals, and terminate the iterative process of the algorithm.

[0044] Preferably, in step 2, the method for randomly generating the initial population is: randomly generating a series of HRC parallel disassembly sequences based on the disassembly priority relationship constraints between the tasks, specifically:

[0045] 2.1.1) According to PD, find all tasks whose immediate predecessors are empty or have been assigned, that is, find the tasks corresponding to the sum of all column elements in PD being 0, and form the task set CS to be assigned;

[0046] 2.1.2) Randomly select a task T in CS and assign it to the current position sequence j of the current individual Pop_i;

[0047] 2.1.3) If the component corresponding to task T is a hazardous component, task T will be assigned to a robot for operation; if the component corresponding to task T is a component that needs to be remanufactured, task T will be assigned to a worker for operation; otherwise, task T will be randomly assigned to a certain type of task operator;

[0048] 2.1.4) Set the row element of task T in PD to 0 and the column element to 1.

[0049] Preferably, in step 2, the method for calculating the objective function value of each population individual is a decoding process, which sequentially assigns each disassembly task of the HRC disassembly sequence to each task operator of each HRC disassembly workstation, and then calculates the objective function value of each optimization target of the HRC disassembly sequence according to the task allocation scheme, specifically:

[0050] 2.1) Open the first workstation of the HRC disassembly line, k = 1;

[0051] 2.2) Let the total working time of the worker at the current workstation k be 0, that is, WT_k_H = 0, and let the remaining working time of the worker at the current workstation k be CT, that is, RT_k_H = CT; let the total working time of the robot at the current workstation k be 0, that is, WT_k_R = 0, and let the remaining working time of the robot at the current workstation k be CT, that is, RT_k_R = CT;

[0052] 2.3) If the operator of disassembly task j is a worker, proceed to the next step, otherwise, proceed to step 2.7);

[0053] 2.4) Calculate the earliest time FT_k_H at which disassembly task j can start execution, and proceed to the next step;

[0054] 2.5) If the operation time t_j_H of disassembly task j is greater than min((CT-FT_k_H), RT_k_H), then open a new workstation, set k=k+1, and repeat step 2.2); otherwise, proceed to the next step;

[0055] 2.6) If the operation time of disassembly task j is t_j_H <= min((CT-FT_k_H), RT_k_H), assign disassembly task j to the current workstation k; update the total operation time of the workers at the current workstation k, i.e., let WT_k_H = WT_k_H + t_j_H, and update the remaining operation time of the workers at the current workstation k, i.e., let RT_k_H = CT-WT_k_H;

[0056] 2.7) If the operator of disassembly task j is a robot, calculate the earliest time FT_k_R when disassembly task j can start to execute and proceed to the next step;

[0057] 2.8) If the operation time t_j_R of disassembly task j>min((CT-FT_k_R),RT_k_R), open a new workstation, set k=k+1, and repeat step 2.2); otherwise proceed to the next step;

[0058] 2.9) If the operation time of disassembly task j is t_j_R<=min((CT-FT_k_R),RT_k_R), assign disassembly task j to the current workstation k, and update the total operation time of the robot at the current workstation k, that is, let WT_k_R=WT_k_R+t_j_R, and at the same time update the remaining operation time of the robot at the current workstation k, that is, let RT_k_R=CT-WT_k_R.

[0059] Preferably, in step 4, specifically:

[0060] If the i-th population individual X cis a student individual, then a teacher individual X is randomly selected from the selected Pareto non-dominated individuals Q_pop p , in X p Randomly select two sequence positions c and d in X; p The order of the corresponding tasks in X c The task between two sequence positions c and d in is based on X p Rearrange the order of the corresponding tasks in order to generate individual X n .

[0061] Preferably, in step 5, the disturbance strategy is specifically:

[0062] For each disturbance, any disassembly task corresponding to the remaining parts in the current HRC disassembly task sequence is randomly selected. If the operator corresponding to the selected disassembly task is a worker, it is transformed into a robot; if the operator corresponding to the selected disassembly task is a robot, it is transformed into a worker.

[0063] Preferably, in step 6, the local search operation based on the single-point insertion operation driven by the greedy mechanism is introduced as follows:

[0064] 6.1) In the current population individual X c Randomly select a task from the

[0065] 6.2) Find the disassembly task immediately preceding and immediately following the randomly selected task;

[0066] 6.3) Determine the sequence position where the randomly selected disassembly task can be inserted;

[0067] 6.4) Insert the randomly selected disassembly task into any insertable sequence position to generate a new individual X n ;

[0068] 6.5) For X n and X c Perform Pareto comparison screening operations;

[0069] 6.6) If X n p X c , X c Update to X n ;

[0070] In step 7, the local search strategy based on the two-point exchange operation is:

[0071] 7.1) In X c Randomly select two disassembly tasks, and respectively determine the preceding disassembly task and the succeeding disassembly task of the two selected disassembly tasks;

[0072] 7.2) If the exchange of the two randomly selected disassembly tasks does not violate the disassembly priority relationship, then, for X c Perform a two-point exchange operation to generate a new individual X n Otherwise, in X c Randomly reselect two disassembly tasks;

[0073] 7.3) For X n and X c Perform Pareto comparison screening operations;

[0074] 7.4) If X n p X c , X c Update to X n .

[0075] The present invention proposes a human-machine collaborative disassembly line balancing problem that takes into account the remanufacturing demand characteristics and hazard characteristics of parts. According to the remanufacturing demand characteristics and hazard characteristics of the parts corresponding to each disassembly task, the disassembly tasks are divided into demand tasks, dangerous tasks and general tasks, and a disassembly task-operator mapping constraint mechanism is established. A mixed integer programming model of the studied problem is constructed with the optimization objectives of minimizing the number of workstations, minimizing the idle time balance index of the task operator, minimizing the demand index and minimizing the hazard index. An improved teaching and learning optimization algorithm is proposed for solving this problem. Two-point crossover operations are introduced in the teaching stage, and single-point insertion and two-point exchange operations are introduced in the learning stage. In addition, a perturbation strategy based on randomly transforming task operators is embedded in the teaching and learning stages to improve the global optimization performance of the MTLO algorithm.

[0076] Based on the remanufacturing demand and hazard characteristics of each component of waste products, the present invention establishes a mapping constraint mechanism between disassembly tasks and task operators, and assigns disassembly operators to various tasks according to the mapping constraint mechanism. A MIP model with the optimization objectives of minimizing the number of workstations, minimizing the idle time balance index of task operators, minimizing the demand index, and minimizing the hazard index is constructed for the proposed HRC-DLBP problem, which can better solve the multi-objective human-machine collaborative disassembly line balance problem. From the perspective of economy and environment, the present invention assigns the disassembly tasks corresponding to the components with hazardous characteristics to robots as much as possible to avoid threatening the health of workers. On the other hand, the disassembly tasks corresponding to the components that can be directly recycled or reused after remanufacturing should be assigned to workers as much as possible to ensure that these components can be dismantled intact, thereby creating value for disassembly enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 It is a flow chart of a method for solving the multi-objective human-machine collaborative disassembly line balance problem in an embodiment;

[0078] Figure 2 This is a schematic diagram of a machine cooperative disassembly line in an embodiment;

[0079] Figure 3 Schematic diagram of coding of HRC-DLBP in the embodiment;

[0080] Figure 4 A schematic diagram of a two-point crossover operation in an embodiment;

[0081] Figure 5 Schematic diagram of a perturbation strategy based on random transformation of task operators in an embodiment;

[0082] FIG6( a ) is a schematic diagram of a single-point insertion operation in an embodiment;

[0083] FIG6( b ) is a schematic diagram of a two-point exchange operation in an embodiment. DETAILED DESCRIPTION

[0084] 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.

[0085] Example

[0086] like Figure 1 As shown, this embodiment provides a multi-objective human-machine collaborative disassembly line balancing problem solution method, which includes the following steps:

[0087] First, a human-robot collaborative disassembly line balancing problem (HRC-DLBP) is proposed, which takes into account the remanufacturing demand and hazard characteristics of parts. Aiming at the task operator allocation problem, a mapping constraint mechanism between disassembly tasks and task operators is established based on the remanufacturing demand and hazard characteristics of each part of waste products.

[0088] Second, a mixed integer programming model is constructed to optimize the problem by minimizing the number of workstations, minimizing the idle time balance index of task operators, and minimizing the demand index and the danger index.

[0089] 3. The solution is obtained through a multi-objective improved teaching optimization MTLO algorithm that integrates an embedded perturbation strategy based on randomly transformed task operators.

[0090] In step 1, a human-machine collaborative disassembly line balancing problem HRC-DLBP considering the demand characteristics and hazard characteristics of component remanufacturing is:

[0091] The schematic diagram of the HRC disassembly line of this embodiment is as follows Figure 2As shown in the figure, there are two types of task operators on the HRC disassembly line: workers and robots. Each workstation on the HRC disassembly line can only be equipped with one worker and one robot to perform disassembly tasks in parallel. In order to solve the operator allocation problem of disassembly tasks in HRC-DLBP, all disassembly tasks are divided into three categories according to the hazards and remanufacturing demand characteristics of the parts corresponding to the disassembly tasks:

[0092] a) Disassembly tasks corresponding to hazardous parts are only allowed to be performed by robots;

[0093] b) The disassembly tasks corresponding to the parts that need to be remanufactured are only allowed to be performed manually;

[0094] c) Other tasks that can be performed by robots or workers;

[0095] Taking into account that workers and robots have their own task execution capabilities and characteristics, a mapping constraint mechanism is constructed to characterize the matching relationship between task operators and disassembly tasks with different characteristic attributes; on this basis, combined with the derivation of the priority relationship constraints among the disassembly tasks of waste products, under the constraint of following the production rhythm, the disassembly tasks are assigned to the task operators at each workstation of the HRC disassembly line, and a balancing scheme for the HRC disassembly line is sought with the minimum number of disassembly workstations, the minimum idle time balance index of the task operators, the minimum danger index and the minimum demand index.

[0096] When solving HRC-DLBP, the following assumptions are made:

[0097] 1) Consider the complete disassembly mode, and the disassembly task corresponding to each component can only be assigned to one task operator to perform independently, that is, there is no task involving human-computer interaction;

[0098] 2) The time it takes for workers and robots to disassemble each component is known;

[0099] 3) The operator's walking time and tool changing time are included in the disassembly operation time;

[0100] 4) The physical implementation details of the disassembly process ensure effective collaboration between workers and robots;

[0101] 5) Ignore all uncertainties and emergency situations in the HRC disassembly production process.

[0102] In step 2, the objective function of the mixed integer programming model is:

[0103]

[0104] Formulas (1)-(4) describe four optimization objectives respectively: minimizing the number of HRC disassembly workstations, minimizing the idle time balance index of task operators, minimizing the demand index, and minimizing the hazard index. From the perspective of disassembly cost, it is necessary to minimize the number of open workstations to minimize the disassembly cost. On the other hand, in order to reduce the number of task operators and make the workload of each operator as full and balanced as possible, the idle time balance index of the task operator should be minimized as much as possible. Considering that some parts with hazardous characteristics may cause damage or contamination to the remaining parts of the waste product during the disassembly process, these parts with hazardous characteristics should be removed as early as possible. In addition, parts with remanufacturing requirements should be removed as early as possible to prevent these parts with remanufacturing requirements from being damaged during the disassembly process.

[0105] The model constraints are:

[0106]

[0107]

[0108] Where i, j represent the disassembly task index; p represents the disassembly task sequence number index; k represents the HRC disassembly workstation index; w represents the task operator, if the task operator is a robot, w = 1, if the task operator is a worker, w = 2; CT represents the production line cycle; t i w represents the operation time required for task operator w to perform disassembly task i; D i represents the demand index. If the parts corresponding to disassembly task i are remanufacturable, then D i =1; otherwise, D i =0; H i represents the hazard index. If the component corresponding to the disassembly task i is hazardous, then H i =1; otherwise, H i =0; t bi represents the time when task i starts to execute; M represents a positive number; PD represents the disassembly priority relationship matrix PD = [pd ij ] I×I , if task i is followed by task j, pd ij =1; otherwise, pd ij =0; Z k Represents a binary variable. If HRC disassembly station k is turned on, Z k =1; otherwise, Z k =0; U k w represents a binary variable. If task operator w is assigned to workstation k, U k w =1; otherwise Uk w =0;x ik w represents a binary variable, if task i is assigned to task operator w in workstation k, then x ik w =1; otherwise x ik w =0;Q ijk w represents a binary variable. If disassembly tasks i and j are assigned to the same task operator w in the same workstation k, and disassembly task i is executed before task j, then Q ijk w =1; otherwise, Q ijk w =0;s ip represents a binary variable. If the disassembly sequence number of disassembly task i is p, s ip =1; otherwise s ip =0;

[0109] Formula (5) indicates that any disassembly task must be assigned and can only be assigned to a task operator in a certain workstation; Formula (6) indicates that the disassembly task corresponding to hazardous parts must be assigned to robot disassembly; Formula (7) ensures that the disassembly task corresponding to parts with remanufacturing requirements must be assigned to workers for disassembly; Formula (8) indicates that if a workstation is open, the workstation must be equipped with at least one type of task operator, and can be equipped with at most two different types of task operators; Formulas (9)-(10) indicate that if a workstation is equipped with a certain type of task operator, there must be at least one disassembly task assigned to the task operator, and the number of tasks assigned to the task operator cannot exceed the total number of disassembly tasks; Formulas (11)-(12) indicate that if a human-machine collaborative disassembly workstation is open, it must be guaranteed that at least one disassembly task is assigned to the workstation, and the total number of tasks assigned to the workstation cannot exceed the total number of disassembly tasks; Formula (13) indicates that the workstations are opened in sequence and there are no empty workstations;

[0110] Formula (14) indicates that if disassembly task i is the immediate predecessor of task j, the start execution time of disassembly task j shall not be earlier than the execution completion time of disassembly task i; Formula (15) indicates that if any two different disassembly tasks i and j are assigned to the same task operator of a workstation, and task i is executed before task j, the start execution time of disassembly task j shall not be earlier than the completion time of task i; Formulas (16)-(17) indicate that if two disassembly tasks i and j are assigned to the same task operator in the same workstation, the two disassembly tasks i and j cannot be operated in parallel at the same time; Formula (18) indicates that if two different disassembly tasks i and j are assigned to the same task operator of the workstation, and task i is executed before task j, the start execution time of task j must be greater than the sum of the start time of the corresponding workstation and the disassembly time of task i; Formula (19) indicates that any task i must be disassembled within the cycle time; Formula (20) stipulates that the start time of any disassembly task i is a positive real number.

[0111] In step three, the Teaching and Learning Optimization (TLO) algorithm is an intelligent optimization algorithm. Because of its advantages such as simple concept, few parameters, and simplified initial setting process of algorithm optimization, it has been applied to solving discrete combinatorial optimization problems such as automobile assembly lines, multi-workshop collaborative scheduling, and parallel machine scheduling problems, and has shown excellent solving performance. Considering that the HRC-DLBP studied in this embodiment is a typical NP-complete problem, and the problem has the dual characteristics of discrete combinatorial optimization and multi-objective optimization, this embodiment proposes an improved teaching optimization algorithm (Modified Teaching and Learning Optimization, MTLO), and designs a two-layer encoding strategy containing dual information of task-operator mapping and human-machine collaborative parallel disassembly sequence according to the problem-solving characteristics, establishes a teaching mechanism based on genetic crossover operation and a mutual learning mechanism based on single-point mutation operation, and accelerates the convergence speed of the algorithm while improving the global optimization performance of the algorithm. In the self-learning stage, a local iterative optimization mechanism based on exchange operation is introduced, and a perturbation strategy based on random transformation of task operators is designed in the teaching, mutual learning, and self-learning stages, so as to effectively enhance the local optimization performance of the algorithm.

[0112] The specific implementation steps of the MTLO algorithm are as follows:

[0113] Step 1: Initialize the algorithm parameters, including the number of individuals in the population N, the number of iterations Iter, the number of disturbances rd, the learning phase 1 parameter Num_ls1 and the learning phase 2 parameter Lk; input the problem parameters: the operation time of each disassembly task, the operation priority relationship between disassembly tasks, and the scheduled HRC disassembly line production rhythm;

[0114] Step 2: Randomly generate the initial population, calculate the objective function value of each individual in the population, and screen the Pareto non-dominated individuals on this basis;

[0115] Step 3: Set the iteration count to 1, that is, iter = 1;

[0116] Step 4: During the teaching phase, for each student individual, a random neighborhood search operation of a two-point crossover operation is performed to update the current student individual;

[0117] Step 5: Implement the rd non-repetitive perturbation strategy for each student individual obtained in step 4 to update the current student individual;

[0118] Step 6: In learning phase 1, perform a local search operation based on a single-point insertion operation driven by a greedy mechanism on each student individual obtained in step 5 to generate Num_ls1 neighborhood individuals, and perform a Pareto comparison between the current student individual and the newly generated neighborhood individuals; use the selected Pareto non-dominated individuals to update the current student individual; then implement the rd non-repeated perturbation strategy for each student individual to update the current student individual;

[0119] Step 7: In the learning phase 2, a local search strategy of two-point exchange operation is performed on each student individual obtained in step 6, and the knowledge level of these student individuals is continuously updated; on this basis, a perturbation strategy of rd times without repetition is implemented for each student individual to update the current student individual;

[0120] Step 8: Update the current Pareto non-dominated individuals by performing Pareto comparison on all current student individuals;

[0121] Step 9: iter=iter+1, if iter<=Iter, return to step 4; otherwise, go to step 10;

[0122] Step 10: Screen and obtain the final Pareto non-dominated individuals, and terminate the iterative process of the algorithm.

[0123] In step 2, before using the proposed MTLO algorithm to solve the HRC-DLBP problem, the solution to the problem must first be encoded. Considering the multi-level coupled optimization solution characteristics of the HRC-DLBP problem, this embodiment designs a two-layer encoding strategy that contains dual information of task-operator mapping and HRC parallel disassembly sequence. Specifically, each individual of the MTLO algorithm corresponds to an HRC parallel disassembly sequence, which contains two layers of information: the task execution order and the operator corresponding to each task. Figure 3As shown in the figure, a waste product with 10 disassembly tasks is used as an example to illustrate the coding method of HRC-DLBP. The disassembly task priority diagram of this product is shown in Figure 3 It is given in the disassembly priority relationship diagram corresponding to the product, the letter D next to the task indicates that the component corresponding to the task has a remanufacturing demand, which is represented by a dark circle, and the letter H next to the task indicates that the component corresponding to the task is hazardous, which is represented by a light circle. When a component has both remanufacturing demand and hazardous characteristics, the component will be regarded as a hazardous component, and its remanufacturing demand characteristics will be ignored. Population initialization is to randomly generate a series of HRC parallel disassembly sequences based on the disassembly priority relationship constraints between various tasks. The method of randomly generating the initial population is: randomly generate a series of HRC parallel disassembly sequences based on the disassembly priority relationship constraints between various tasks, specifically:

[0124] 2.1.1) According to PD, find all tasks whose immediate predecessors are empty or have been assigned, that is, find the tasks corresponding to the sum of all column elements in PD being 0, and form the task set CS to be assigned;

[0125] 2.1.2) Randomly select a task T in CS and assign it to the current position sequence j of the current individual Pop_i;

[0126] 2.1.3) If the component corresponding to task T is a hazardous component, task T will be assigned to a robot for operation; if the component corresponding to task T is a component that needs to be remanufactured, task T will be assigned to a worker for operation; otherwise, task T will be randomly assigned to a certain type of task operator;

[0127] 2.1.4) Set the row element of task T in PD to 0 and the column element to 1.

[0128] The following pseudo code shows the steps to generate the initial population:

[0129]

[0130] In step 2, the method for calculating the objective function value of each population individual is the decoding process. Decoding is to assign the various disassembly tasks of the HRC disassembly sequence to the various task operators of each HRC disassembly workstation in sequence under given constraints, and then calculate the objective function values ​​of the HRC disassembly sequence corresponding to each optimization goal according to the task allocation plan, specifically:

[0131] 2.1) Open the first workstation of the HRC disassembly line, k = 1;

[0132] 2.2) Let the total working time of the worker at the current workstation k be 0, that is, WT_k_H = 0, and let the remaining working time of the worker at the current workstation k be CT, that is, RT_k_H = CT; let the total working time of the robot at the current workstation k be 0, that is, WT_k_R = 0, and let the remaining working time of the robot at the current workstation k be CT, that is, RT_k_R = CT;

[0133] 2.3) If the operator of disassembly task j is a worker, proceed to the next step, otherwise, proceed to step 2.7);

[0134] 2.4) Calculate the earliest time FT_k_H at which disassembly task j can start execution, and proceed to the next step;

[0135] 2.5) If the operation time t_j_H of disassembly task j is greater than min((CT-FT_k_H), RT_k_H), then open a new workstation, set k=k+1, and repeat step 2.2); otherwise, proceed to the next step;

[0136] 2.6) If the operation time of disassembly task j is t_j_H <= min((CT-FT_k_H), RT_k_H), assign disassembly task j to the current workstation k; update the total operation time of the workers at the current workstation k, i.e., let WT_k_H = WT_k_H + t_j_H, and update the remaining operation time of the workers at the current workstation k, i.e., let RT_k_H = CT-WT_k_H;

[0137] 2.7) If the operator of disassembly task j is a robot, calculate the earliest time FT_k_R when disassembly task j can start to execute and proceed to the next step;

[0138] 2.8) If the operation time t_j_R of disassembly task j>min((CT-FT_k_R),RT_k_R), open a new workstation, set k=k+1, and repeat step 2.2); otherwise proceed to the next step;

[0139] 2.9) If the operation time of disassembly task j is t_j_R<=min((CT-FT_k_R),RT_k_R), assign disassembly task j to the current workstation k, and update the total operation time of the robot at the current workstation k, that is, let WT_k_R=WT_k_R+t_j_R, and at the same time update the remaining operation time of the robot at the current workstation k, that is, let RT_k_R=CT-WT_k_R.

[0140] The following pseudo code gives the specific decoding process:

[0141]

[0142]

[0143] In step 4, during the teaching phase of the MTLO algorithm, students (dominated population individuals) learn from teachers (non-dominated population individuals) to improve themselves. The teacher continuously improves his knowledge level in each algorithm iteration, thus forming a virtuous cycle in which the student's knowledge level continues to improve. The idea of ​​this process is shown in formula (21):

[0144]

[0145] Among them, new student and student represent the individual students whose knowledge level is improved by asking the teacher and the individual students whose knowledge level is improved, respectively. Teacher represents the Pareto non-dominated individual in the current algorithm iteration. r is a random number in the interval (0,1) for the learning step size. TF represents the teaching factor. Mean is the average of the knowledge level of all individual students. Based on the idea of ​​the above teaching stage and combined with the solution characteristics of the HRC-DLBP problem, a two-point crossover operation is designed to enable each individual student in the population to interact with the optimal individual teacher to effectively improve their knowledge level. The specific operation process is as follows: Figure 4 shown.

[0146] Specifically:

[0147] If the i-th population individual X c is a student individual, then a teacher individual X is randomly selected from the selected Pareto non-dominated individuals Q_pop p , in X p Randomly select two sequence positions c and d in X; p The order of the corresponding tasks in X c The task between two sequence positions c and d in is based on X p Rearrange the order of the corresponding tasks in order to generate individual X n .

[0148] The following pseudo code gives the specific teaching process:

[0149]

[0150]

[0151] In step 5, the disturbance strategy is as follows:

[0152] In the HRC disassembly line, except for the hazardous parts that must be assigned to robots and the parts with remanufacturing requirements that must be assigned to workers, the disassembly tasks corresponding to the remaining parts will be randomly assigned to robots or workers. Different operator task allocation schemes will affect the balance of the HRC disassembly line. Based on this, a method is proposed as follows Figure 5 Specifically, for each disturbance, any disassembly task (such as Figure 5 If the operator corresponding to the selected disassembly task is a worker, it is transformed into a robot; if the operator corresponding to the selected disassembly task is a robot, it is transformed into a worker (such as Figure 5 As shown, the operator corresponding to Task 5 is changed from a robot to a worker).

[0153] In steps 6 and 7, the learning phase includes two sub-phases: the mutual learning phase and the self-learning phase. In the mutual learning phase, students learn from each other and continuously improve themselves. In the self-learning phase, students review and consolidate their knowledge according to their own situation. According to the idea of ​​the learning phase and combined with the solution characteristics of the HRC-DLBP problem, a two-phase learning strategy is proposed to improve the global convergence performance of the MTLO algorithm. In the first phase of the learning strategy, a local search strategy based on single-point insertion operation driven by a greedy mechanism is introduced, as shown in Figure 6(a). In the second phase, a local search strategy based on two-point exchange operation is adopted, as shown in Figure 6(b).

[0154] The local search operation based on single-point insertion operation driven by the greedy mechanism is as follows:

[0155] 6.1) In the current population individual X c Randomly select a task from the

[0156] 6.2) Find the disassembly task immediately preceding and immediately following the randomly selected task;

[0157] 6.3) Determine the sequence position where the randomly selected disassembly task can be inserted;

[0158] 6.4) Insert the randomly selected disassembly task into any insertable sequence position to generate a new individual X n ;

[0159] 6.5) For X n and X c Perform Pareto comparison screening operations;

[0160] 6.6) If X np X c , X c Update to X n ;

[0161] The local search strategy based on the two-point exchange operation is:

[0162] 7.1) In X c Randomly select two disassembly tasks, and respectively determine the preceding disassembly task and the succeeding disassembly task of the two selected disassembly tasks;

[0163] 7.2) If the exchange of the two randomly selected disassembly tasks does not violate the disassembly priority relationship, then, for X c Perform a two-point exchange operation to generate a new individual X n Otherwise, in X c Randomly reselect two disassembly tasks;

[0164] 7.3) For X n and X c Perform Pareto comparison screening operations;

[0165] 7.4) If X n p X c , X c Update to X n .

[0166] The pseudo code for the two-stage learning strategy is as follows:

[0167]

[0168]

[0169] Algorithm Verification

[0170] In order to verify the correctness of the multi-objective MIP model built for the HRC-DLBP problem and the effectiveness of the proposed MTLO algorithm, a laptop disassembly example with eight disassembly tasks was introduced and modified.

[22] . The relevant data information of this example, including the disassembly operation time of each task (worker / robot), the demand and hazard index of the parts corresponding to each task, and the job priority relationship between disassembly tasks are listed in Table 1. Among them, D and H represent the demand index and hazard index of the parts corresponding to each task, respectively. If the parts corresponding to the disassembly task have remanufacturing demand, the D index of the task is set to 1, otherwise, the D index is set to 0; if the parts corresponding to the disassembly task are hazardous parts, the H index of the task is set to 1, otherwise the H index is set to 0. In addition, the production cycle time of this laptop disassembly example is set to 40s. Based on LINGO 11.0 software, a MIP model of the HRC-DLBP problem for this example is established, and the optimal solution of each objective is obtained. At the same time, the proposed MTLO algorithm is applied to solve this example. The MTLO algorithm was written and implemented on MATLAB R2020a software. The operating environment of all software is: Windows 10 system, Inter(R) Core(TM) i5-9400 CPU@2.90GHz, 16GB memory. The parameter values ​​of the MTLO algorithm are set as follows: Iter=50, N=50, rd=3, Num_ls1=2, Lk=3.

[0171] Table 1. Relevant data of the laptop disassembly example

[0172]

[0173] Table 2 shows the solution results of the LINGO software and the MTLO algorithm for the laptop disassembly example. As can be seen from Table 2, the proposed MTLO algorithm obtains a non-inferior solution for the example, and the solution results of the four optimization objectives corresponding to the non-inferior solution are all optimal solutions. In addition, the solution time of the MTLO algorithm is only 2.61 seconds, while LINGO needs 145 seconds to obtain the optimal solution of the objective F2. Compared with LINGO, the proposed MTLO algorithm can obtain the optimal solutions of all four objectives at one time in a shorter time, which verifies the correctness of the constructed MIP model and the effectiveness of the MTLO algorithm.

[0174] Table 2 Calculation results of MTLO algorithm and LINGO for laptop disassembly example

[0175]

[0176] In order to further verify the correctness of the constructed multi-objective MIP model and the effectiveness of the proposed MTLO algorithm, a disassembly example containing 10 tasks is introduced and modified. The relevant data information of the modified example is shown in Table 3. The production cycle time of the example is set to 40 seconds. LINGO software and MTLO algorithm are used to solve the example. The parameter values ​​of the proposed MTLO algorithm are set as follows: Iter = 100, N = 50, rd = 5, Num_ls1 = 3, Lk = 10. Table 4 gives the solution results of LINGO and MTLO algorithms for the example. As shown in Table 4, the MTLO algorithm obtains a Pareto non-inferior solution set containing 14 non-inferior solutions. Among them, non-inferior solutions 2, 4, 7 and 12 obtain the optimal value 2 of objective F1; non-inferior solutions 7 and 12 obtain the optimal value 9 of objective F2; non-inferior solutions 6, 8 and 11 obtain the optimal value 3 of objective F3; non-inferior solutions 1 and 3 obtain the optimal value 1 of objective F4. In addition, the MTLO algorithm can obtain the optimal values ​​of all targets at once within 8.1 seconds, while LINGO takes 947 seconds to calculate the optimal value of target F2. This further verifies the correctness of the MIP model established for HRC-DLBP in this embodiment and the effectiveness of the proposed MTLO algorithm.

[0177] Table 4 contains the relevant data of the 10 task disassembly examples.

[0178]

[0179]

[0180] Table 5 Solution results of LINGO and MTLO algorithms for the disassembly example containing 10 tasks

[0181]

[0182] This embodiment proposes a human-machine collaborative disassembly line balancing problem that takes into account the remanufacturing demand characteristics and hazard characteristics of parts. According to the remanufacturing demand characteristics and hazard characteristics of the parts corresponding to each disassembly task, the disassembly tasks are divided into demand tasks, dangerous tasks and general tasks, and a disassembly task-operator mapping constraint mechanism is established. A mixed integer programming model of the studied problem is constructed with the optimization objectives of minimizing the number of workstations, minimizing the task operator idle time balance index, minimizing the demand index and minimizing the hazard index. An improved teaching and learning optimization algorithm is proposed for solving this problem. Two-point crossover operations are introduced in the teaching stage, and single-point insertion and two-point exchange operations are introduced in the learning stage. In addition, a perturbation strategy based on randomly transforming task operators is embedded in the teaching and learning stages to improve the global optimization performance of the MTLO algorithm.

[0183] 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 multi-objective human-machine collaborative disassembly line balance problem solution method, characterized by: The following steps are involved: First, a human-robot collaborative disassembly line balancing problem (HRC-DLBP) is proposed, which takes into account the remanufacturing demand and hazard characteristics of parts. Aiming at the task operator allocation problem, a mapping constraint mechanism between disassembly tasks and task operators is established based on the remanufacturing demand and hazard characteristics of each part of waste products. Second, a mixed integer programming model is constructed to optimize the problem by minimizing the number of workstations, minimizing the idle time balance index of task operators, and minimizing the demand index and the danger index.

3. The solution is obtained through a multi-objective improved teaching optimization MTLO algorithm that integrates an embedded perturbation strategy based on randomly transformed task operators.

2. A multi-objective human-machine collaborative disassembly line balancing problem solving method according to claim 1, characterized in that: In step 1, a human-machine collaborative disassembly line balancing problem HRC-DLBP considering the demand characteristics and hazard characteristics of component remanufacturing is: There are two types of task operators on the HRC disassembly line: workers and robots. Each workstation on the HRC disassembly line can only be equipped with one worker and one robot to perform disassembly tasks in parallel. In order to solve the operator allocation problem of disassembly tasks in HRC-DLBP, all disassembly tasks are divided into three categories according to the hazards and remanufacturing demand characteristics of the parts corresponding to the disassembly tasks: a) Disassembly tasks corresponding to hazardous parts are only allowed to be performed by robots; b) The disassembly tasks corresponding to the parts that need to be remanufactured are only allowed to be performed manually; c) Other tasks that can be performed by robots or workers; Taking into account that workers and robots have their own task execution capabilities and characteristics, a mapping constraint mechanism is constructed to characterize the matching relationship between task operators and disassembly tasks with different characteristic attributes; on this basis, combined with the derivation of the priority relationship constraints among the disassembly tasks of waste products, under the constraint of following the production rhythm, the disassembly tasks are assigned to the task operators at each workstation of the HRC disassembly line, and a balancing scheme for the HRC disassembly line is sought with the minimum number of disassembly workstations, the minimum idle time balance index of the task operators, the minimum danger index and the minimum demand index.

3. A multi-objective human-machine collaborative disassembly line balancing problem solving method according to claim 2, characterized in that: When solving HRC-DLBP, the following assumptions are made: 1) Consider the complete disassembly mode, and the disassembly task corresponding to each component can only be assigned to one task operator to perform independently, that is, there is no task involving human-computer interaction; 2) The time it takes for workers and robots to disassemble each component is known; 3) The operator's walking time and tool changing time are included in the disassembly operation time; 4) The physical implementation details of the disassembly process ensure effective collaboration between workers and robots; 5) Ignore all uncertainties and emergency situations in the HRC disassembly production process.

4. A multi-objective human-machine collaborative disassembly line balancing problem solving method according to claim 3, characterized in that: In step 2, the objective function of the mixed integer programming model is: Formulas (1)-(4) describe four optimization objectives respectively: minimizing the number of HRC disassembly workstations, minimizing the idle time balance index of task operators, minimizing the demand index, and minimizing the hazard index; The model constraints are: Where i, j represent the disassembly task index; p represents the disassembly task sequence number index; k represents the HRC disassembly workstation index; w represents the task operator, if the task operator is a robot, w = 1, if the task operator is a worker, w = 2; CT represents the production line cycle; t i w represents the operation time required for task operator w to perform disassembly task i; D i represents the demand index. If the parts corresponding to disassembly task i are remanufacturable, then D i =1; otherwise, D i =0; H i represents the hazard index. If the component corresponding to the disassembly task i is hazardous, then H i =1; otherwise, H i =0; t bi represents the time when task i starts to execute; M represents a positive number; PD represents the disassembly priority relationship matrix PD = [pd ij ] I×I , if task i is followed by task j, pd ij =1; otherwise, pd ij =0; Z k Represents a binary variable. If HRC disassembly station k is turned on, Z k =1; otherwise, Z k =0; U k w represents a binary variable. If task operator w is assigned to workstation k, U k w =1; otherwise U k w =0;x ik w represents a binary variable, if task i is assigned to task operator w in workstation k, then x ik w =1; otherwise x ik w =0;Q ijk w represents a binary variable. If disassembly tasks i and j are assigned to the same task operator w in the same workstation k, and disassembly task i is executed before task j, then Q ijk w =1; otherwise, Q ijk w =0;s ip represents a binary variable. If the disassembly sequence number of disassembly task i is p, s ip =1; otherwise s ip =0; Formula (5) indicates that any disassembly task must be assigned and can only be assigned to a task operator in a certain workstation; Formula (6) indicates that the disassembly task corresponding to hazardous parts must be assigned to robot disassembly; Formula (7) ensures that the disassembly task corresponding to parts with remanufacturing requirements must be assigned to workers for disassembly; Formula (8) indicates that if a workstation is open, the workstation must be equipped with at least one type of task operator, and can be equipped with at most two different types of task operators; Formulas (9)-(10) indicate that if a workstation is equipped with a certain type of task operator, there must be at least one disassembly task assigned to the task operator, and the number of tasks assigned to the task operator cannot exceed the total number of disassembly tasks; Formulas (11)-(12) indicate that if a human-machine collaborative disassembly workstation is open, it must be guaranteed that at least one disassembly task is assigned to the workstation, and the total number of tasks assigned to the workstation cannot exceed the total number of disassembly tasks; Formula (13) indicates that the workstations are opened in sequence and there are no empty workstations; Formula (14) indicates that if disassembly task i is the immediate predecessor of task j, the start execution time of disassembly task j shall not be earlier than the execution completion time of disassembly task i; Formula (15) indicates that if any two different disassembly tasks i and j are assigned to the same task operator of a workstation, and task i is executed before task j, the start execution time of disassembly task j shall not be earlier than the completion time of task i; Formulas (16)-(17) indicate that if two disassembly tasks i and j are assigned to the same task operator in the same workstation, the two disassembly tasks i and j cannot be operated in parallel at the same time; Formula (18) indicates that if two different disassembly tasks i and j are assigned to the same task operator of the workstation, and task i is executed before task j, the start execution time of task j must be greater than the sum of the start time of the corresponding workstation and the disassembly time of task i; Formula (19) indicates that any task i must be disassembled within the cycle time; Formula (20) stipulates that the start time of any disassembly task i is a positive real number.

5. The multi-objective human-machine collaborative disassembly line balancing problem solving method according to claim 4 is characterized by: In step 3, the specific implementation steps of the MTLO algorithm are as follows: Step 1: Initialize the algorithm parameters, including the number of individuals in the population N, the number of iterations Iter, the number of disturbances rd, the learning phase 1 parameter Num_ls1 and the learning phase 2 parameter Lk; input the problem parameters: the operation time of each disassembly task, the operation priority relationship between disassembly tasks, and the scheduled HRC disassembly line production rhythm; Step 2: Randomly generate the initial population, calculate the objective function value of each individual in the population, and screen the Pareto non-dominated individuals on this basis; Step 3: Set the iteration count to 1, that is, iter = 1; Step 4: During the teaching phase, for each student individual, a random neighborhood search operation of a two-point crossover operation is performed to update the current student individual; Step 5: Implement the rd non-repetitive perturbation strategy for each student individual obtained in step 4 to update the current student individual; Step 6: In learning phase 1, perform a local search operation based on a single-point insertion operation driven by a greedy mechanism on each student individual obtained in step 5 to generate Num_ls1 neighborhood individuals, and perform a Pareto comparison between the current student individual and the newly generated neighborhood individuals; use the selected Pareto non-dominated individuals to update the current student individual; then implement the rd non-repeated perturbation strategy for each student individual to update the current student individual; Step 7: In the learning phase 2, a local search strategy of two-point exchange operation is performed on each student individual obtained in step 6, and the knowledge level of these student individuals is continuously updated; on this basis, a perturbation strategy of rd times without repetition is implemented for each student individual to update the current student individual; Step 8: Update the current Pareto non-dominated individuals by performing Pareto comparison on all current student individuals; Step 9: iter=iter+1, if iter<=Iter, return to step 4; otherwise, go to step 10; Step 10: Screen and obtain the final Pareto non-dominated individuals, and terminate the iterative process of the algorithm.

6. A multi-objective human-machine collaborative disassembly line balancing problem solving method according to claim 5, characterized in that: In step 2, the method of randomly generating the initial population is: randomly generating a series of HRC parallel disassembly sequences based on the disassembly priority relationship constraints between the tasks, specifically: 2.1.1) According to PD, find all tasks whose immediate predecessors are empty or have been assigned, that is, find the tasks corresponding to the sum of all column elements in PD being 0, and form the task set CS to be assigned; 2.1.2) Randomly select a task T in CS and assign it to the current position sequence j of the current individual Pop_i; 2.1.3) If the component corresponding to task T is a hazardous component, task T will be assigned to a robot for operation; if the component corresponding to task T is a component that needs to be remanufactured, task T will be assigned to a worker for operation; otherwise, task T will be randomly assigned to a certain type of task operator; 2.1.4) Set the row element of task T in PD to 0 and the column element to 1.

7. A multi-objective human-machine collaborative disassembly line balancing problem solving method according to claim 6, characterized in that: In step 2, the method for calculating the objective function value of each population individual is the decoding process, which sequentially assigns each disassembly task of the HRC disassembly sequence to each task operator of each HRC disassembly workstation, and then calculates the objective function value of each optimization target of the HRC disassembly sequence according to the task allocation scheme, specifically: 2.1) Open the first workstation of the HRC disassembly line, k = 1; 2.2) Let the total working time of the worker at the current workstation k be 0, that is, WT_k_H = 0, and let the remaining working time of the worker at the current workstation k be CT, that is, RT_k_H = CT; let the total working time of the robot at the current workstation k be 0, that is, WT_k_R = 0, and let the remaining working time of the robot at the current workstation k be CT, that is, RT_k_R = CT; 2.3) If the operator of disassembly task j is a worker, proceed to the next step, otherwise, proceed to step 2.7); 2.4) Calculate the earliest time FT_k_H at which disassembly task j can start execution, and proceed to the next step; 2.5) If the operation time t_j_H of disassembly task j is greater than min((CT-FT_k_H), RT_k_H), then open a new workstation, set k=k+1, and repeat step 2.2); otherwise, proceed to the next step; 2.6) If the operation time of disassembly task j is t_j_H <= min((CT-FT_k_H), RT_k_H), assign disassembly task j to the current workstation k; update the total operation time of the workers at the current workstation k, i.e., let WT_k_H = WT_k_H + t_j_H, and update the remaining operation time of the workers at the current workstation k, i.e., let RT_k_H = CT-WT_k_H; 2.7) If the operator of disassembly task j is a robot, calculate the earliest time FT_k_R at which disassembly task j can start to execute and proceed to the next step; 2.8) If the operation time t_j_R of disassembly task j is greater than min((CT-FT_k_R), RT_k_R), then open a new workstation, set k=k+1, and repeat step 2.2); otherwise, proceed to the next step; 2.9) If the operation time of disassembly task j is t_j_R<=min((CT-FT_k_R),RT_k_R), assign disassembly task j to the current workstation k, and update the total operation time of the robot at the current workstation k, that is, let WT_k_R=WT_k_R+t_j_R, and at the same time update the remaining operation time of the robot at the current workstation k, that is, let RT_k_R=CT-WT_k_R.

8. The multi-objective human-machine collaborative disassembly line balance problem solving method according to claim 7 is characterized by: In step 4, specifically: If the i-th population individual X c is a student individual, then a teacher individual X is randomly selected from the selected Pareto non-dominated individuals Q_pop p , in X p Randomly select two sequence positions c and d in X; p The order of the corresponding tasks in X c The task between two sequence positions c and d in is based on X p Rearrange the order of the corresponding tasks in order to generate individual X n .

9. A multi-objective human-machine collaborative disassembly line balancing problem solving method according to claim 8, characterized in that: In step 5, the disturbance strategy is as follows: For each disturbance, randomly select any disassembly task corresponding to the remaining parts in the current HRC disassembly task sequence. If the operator corresponding to the selected disassembly task is a worker, transform it into a robot. If the operator corresponding to the selected disassembly task is a robot, it is transformed into a worker.

10. A multi-objective human-machine collaborative disassembly line balancing problem solving method according to claim 9, characterized in that: In step 6, the local search operation based on the single-point insertion operation driven by the greedy mechanism is introduced as follows: 6.1) In the current population individual X c Randomly select a task from the 6.2) Find the disassembly task immediately preceding and immediately following the randomly selected task; 6.3) Determine the sequence position where the randomly selected disassembly task can be inserted; 6.4) Insert the randomly selected disassembly task into any insertable sequence position to generate a new individual X n ; 6.5) For X n and X c Perform Pareto comparison screening operations; 6.6) If X n p X c , X c Update to X n ; In step 7, the local search strategy based on the two-point exchange operation is: 7.1) In X c Randomly select two disassembly tasks, and respectively determine the preceding disassembly task and the succeeding disassembly task of the two selected disassembly tasks; 7.2) If the exchange of the two randomly selected disassembly tasks does not violate the disassembly priority relationship, then, for X c Perform a two-point exchange operation to generate a new individual X n Otherwise, in X c Two disassembly tasks were randomly selected again; 7.3) For X n and X c Perform Pareto comparison screening operations; 7.4) If X n p X c , X c Update to X n .

Images