Multi-constraint man-machine cooperation U-shaped disassembly line optimization method

Abstract

The invention discloses a multi-constraint man-machine cooperation U-shaped disassembly line optimization method which comprises the following steps: S1, a disassembly model is established, the disassembly model comprises an objective function and constraint conditions, and the objective function is to minimize the number of started workstations, disassembly line idle time and disassembly line cost; s2, generating an initial population based on the disassembly task and the constraint condition; s3, performing crossover, variation and cloning operation on the initial population, updating a better solution based on a simulated annealing algorithm and in combination with Pareto optimum until the current temperature of the simulated annealing algorithm is smaller than a threshold value, modifying a cooling mode of the simulated annealing algorithm into linear cooling, and repeating the previous steps until the current temperature is smaller than the final temperature, and outputting an optimal solution. According to the method, the comprehensive cost of the disassembly line is considered, the method is closer to the actual production condition, and guidance can be better provided for actual production.

Description

Technical Field

[0001] The invention relates to the technical field of disassembly line optimization, and in particular to a multi-constraint human-machine collaborative U-shaped disassembly line optimization method. Background Art

[0002] For a U-shaped disassembly line, its overall structure is U-shaped, including an inlet and an outlet, and an operator can perform disassembly operations at the workstation on the inlet side and the workstation on the outlet side of the U-shaped disassembly line at the same time without changing his position. In the actual operation of the human-machine collaborative disassembly line, in order to prevent interference between the worker operator and the robot operator, the worker operator and the robot operator are usually set to be located in different workstations separately. Since the working time of the worker operator and the robot operator is different, the number of tasks that can be assigned to the workstation cannot be determined when the operator is not determined at the workstation; since different tasks need to be assigned to different operators to perform the work, the operator in the workstation cannot be determined when the task attributes in the workstation are not determined. Such contradictions increase the difficulty of solving the balancing problem of multi-constraint human-machine collaborative disassembly lines.

[0003] At the same time, in the process of joint disassembly by humans and machines, the disassembly cost issue usually needs to be considered: the purchase cost of the machine is high, but the subsequent maintenance cost is relatively low; there is no purchase cost for labor, but the subsequent production cost is high. Therefore, how to formulate different disassembly plans according to production time is not considered in the prior art. Summary of the invention

[0004] In order to solve at least one of the above problems, the present invention proposes a multi-constraint human-machine collaborative U-shaped disassembly line optimization method.

[0005] The technical solution of the present invention is: a multi-constraint human-machine collaborative U-shaped disassembly line optimization method, comprising the following steps:

[0006] S1. Establish a disassembly model, wherein the disassembly model includes an objective function and constraints. The objective function is to minimize the number of open workstations, the idle time of the disassembly line, and the disassembly line cost. The disassembly line cost includes the machine purchase cost and the disassembly line operation cost. The constraints include task allocation constraints, workstation and disassembler allocation constraints, complex task and hazardous task allocation constraints, takt time constraints, predecessor task constraints, and workstation opening constraints.

[0007] S2, generating an initial population based on the disassembly task and constraints;

[0008] S3. Perform crossover, mutation and cloning operations on the initial population, update the better solution based on the simulated annealing algorithm and combined with Pareto optimality, until the current temperature of the simulated annealing algorithm is less than the threshold, modify the cooling method of the simulated annealing algorithm to linear cooling and repeat the above steps until the current temperature is less than the final temperature, and output the optimal solution; the linear cooling method is as follows: T1=Tt, T≤t0, where T1 represents the temperature after cooling, T represents the temperature before cooling, t is a fixed cooling value, and t0 is a threshold.

[0009] Beneficial effect: The method of the present invention takes into account the comprehensive cost of the disassembly line, wherein the comprehensive cost includes the machine purchase cost and the disassembly line operation cost, and the disassembly line operation cost is positively correlated with the working time of the disassembly line. Therefore, the scheme of the present invention takes the comprehensive cost into consideration, which is closer to the actual production situation and is more conducive to guiding the actual production. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is the optimal allocation scheme for the three objectives of the P25 problem in the embodiment of the present invention;

[0011] Figure 2 It is an HV iteration diagram of the method according to the embodiment of the present invention and the existing method. DETAILED DESCRIPTION

[0012] The specific implementation modes of the present invention will be described clearly and completely below with reference to examples and drawings. Obviously, the described examples are only some embodiments of the present invention, rather than all embodiments.

[0013] Example 1

[0014] A multi-constraint human-machine collaborative U-shaped disassembly line optimization method comprises the following steps:

[0015] S1. Establish a disassembly model, wherein the disassembly model includes an objective function and constraints. The objective function is to minimize the number of open workstations, the idle time of the disassembly line, and the disassembly line cost. The disassembly line cost includes the machine purchase cost and the disassembly line operation cost. The constraints include task allocation constraints, workstation and disassembler allocation constraints, complex task and hazardous task allocation constraints, takt time constraints, predecessor task constraints, and workstation opening constraints.

[0016] Specifically, in order to implement the method of this embodiment, reasonable assumptions are made: 1. The disassembled product is single and complete, with no missing parts; 2. The disassembly priority relationship is known, the disassembly task time is known, the disassembly cost is known, and the task attributes are known; 3. The disassembly supply is unlimited during the disassembly line; 4. There is no interference between the workstations during operation, and the disassembly line has no unexpected interruption; 5. The operator's movement time is ignored. These assumptions are conventional assumptions in this field.

[0017] At the same time, in the actual disassembly process, the properties of the internal parts of the end-of-life products are relatively complex, but they can usually be divided into ordinary parts, hazardous parts and complex parts. Among them, hazardous parts refer to parts that are harmful to workers or the environment during the disassembly process. For this type of parts, it is usually necessary to arrange for machines to disassemble them to reduce the impact on workers; complex parts refer to parts that are relatively difficult to disassemble. This type of parts is difficult for machines to disassemble, so it is usually necessary to arrange for workers to disassemble them.

[0018] In this embodiment, the objective function is as follows:

[0019]

[0020] Where F represents the minimization objective; f1 represents the number of open workstations; f2 represents the idle time of the disassembly line; f3 represents the cost of the disassembly line; N is the set of workstations, and n∈N; l represents the operator, when l=1, it means the operator is a worker, and when l=2, it means the operator is a machine; z ln is a binary variable, z ln =1 means the nth workstation is open and the operator is l, otherwise z ln =0; CT represents the cycle time; m represents the import and export sides of the U-shaped disassembly line, where m=1 represents the import side and m=2 represents the export side; represents the disassembly time of task i when the operator is l; is a binary variable, Indicates that task i is assigned to port m of the nth worker station of operator l, otherwise C r represents the purchase cost of a unit machine; D represents the number of days the U-shaped disassembly line is in operation; G represents the total number of parts disassembled per day; C d represents the cost of disassembling each product; represents the cost of disassembling task i for operator l; It represents the standby cost of the workstation when the operator is l.

[0021] For the above objective function, the minimized number of open workstations f1 aims to reduce the operation length and floor space of the disassembly line, thereby reducing costs and improving efficiency; the minimized idle time of the disassembly line f2 aims to balance the load of each workstation and avoid congestion or blockage at each workstation due to uneven task distribution; the minimized disassembly line cost f3 represents the comprehensive cost of the disassembly line, which includes the purchase cost of the machine and the operating cost of the disassembly line, among which the operating cost of the disassembly line is positively correlated with the running time of the disassembly line. Therefore, by minimizing the disassembly line cost, decision makers can weigh the disassembly distribution of machines and workers according to the running time of the disassembly line.

[0022] In this embodiment, the constraints are as follows:

[0023] Task allocation constraints: Where N is the set of workstations, and n∈N; m represents the import and export sides of the U-shaped disassembly line, where m=1 represents the import side and m=2 represents the export side; l represents the operator, when l=1, it means the operator is a worker, and when l=2, it means the operator is a machine; is a binary variable, Indicates that task i is assigned to port m of the nth worker station of operator l, otherwise This constraint means that each task must and can only be assigned to one end of a workstation.

[0024] Workstation and disassembler allocation constraints: In the formula, M is greater than, z ln is a binary variable, z ln =1 means the nth workstation is open and the operator is l, otherwise z ln =0; For this constraint, it means that each workstation can only be assigned one type of operator, either a worker or a machine, and only one can be selected.

[0025] Complex and hazardous task allocation constraints: Where K represents the set of complex tasks; H represents the set of hazardous tasks. This constraint indicates that tasks with complex attributes need to be assigned to workstations where operators are workers, and tasks with hazardous attributes need to be assigned to workstations where operators are machines. This not only increases the disassembly efficiency of tasks with complex attributes, but also greatly reduces the impact of hazardous attribute tasks on the physical and mental health of workers.

[0026] Takt time constraints: Where, CT represents the takt time;

[0027] Predecessor task constraints:

[0028] Where TP ij Represents the task priority relationship. If task i is the AND immediate predecessor of task j, then TP ij =1, if task i is the OR immediate predecessor of task j, then TP ij =-1, otherwise TP ij =0;OR j represents the set of OR immediate predecessor tasks of task j; y ij is a binary variable, y ij =1 indicates that task i is the OR immediate predecessor of task j and task i is completed before task j, otherwise y ij =0; is a binary variable, indicates that task i is assigned to the import of the nth worker station of operator l, otherwise is a binary variable, indicates that task j is assigned to the import of the nth worker station of operator l, otherwise is a binary variable, Indicates that task i is assigned to the exit of the nth worker station of operator l, otherwise is a binary variable, indicates that task j is assigned to the exit of the nth worker station of operator l, otherwise This constraint condition limits the tasks that precede and follow each part in the end-of-life product. For end-of-life products, the disassembly order between the various parts needs to be disassembled according to certain requirements. For example, for a part, it needs to be disassembled before other parts can be disassembled. For such parts, it is usually said that they have an AND priority relationship constraint, that is, the monetary task must be disassembled before the task that follows it. For other parts, if they need to be disassembled, any of the previous tasks must be disassembled before they can be disassembled. For such parts, it is usually said that they have an OR priority relationship constraint.

[0029] Workstation startup constraints: In this constraint, it means that the opening of workstations needs to correspond to tasks, and workstations need to be opened in sequence.

[0030] S2, generating an initial population based on the disassembly task and constraints;

[0031] Specifically, in this embodiment, the initial population is generated based on encoding and decoding operations.

[0032] Wherein, the encoding step includes the following sub-steps:

[0033] Based on the constraints, firstly, the tasks without immediate predecessors are assigned to the entrance side of the disassembly line, and their task sequences are marked as positive numbers. The tasks without future tasks are assigned to the exit side of the disassembly line, and their task sequences are marked as negative numbers. For the U-shaped disassembly line, it includes an entrance side and an exit side, and the workstation is set between the entrance side and the exit side, and a workstation can usually set disassembly tasks on both the entrance side and the exit side. Since the traditional encoding method cannot search the global solution space, in this embodiment, a mixed method of positive and negative numbers is used to encode the task sequence. For example, for task 5, if it is assigned to the entrance side for disassembly, it is numbered 5, and if it is assigned to the exit side for disassembly, it is numbered -5.

[0034] Then the priority relationship of the remaining tasks is updated, and the above operations are repeated until all tasks are assigned, and finally a task sequence is generated.

[0035] Combined with the operator sequence of the workstation, a two-layer code is generated. In this sequence, the operators are numbered, for example, when the operator is a worker, the code is 1, and when the operator is a machine, the code is 2.

[0036] The decoding step comprises the following sub-steps:

[0037] Assign operators to workstations according to the operator sequence and assign tasks until the takt time is reached;

[0038] Check whether there is a conflict between the task and the operator. If there is a conflict, change the operator;

[0039] After changing the operator, if the non-primary task and the operator still conflict, the operator will be changed back to the original operator; after changing the operator, if the primary task and the operator conflict, the operator of the workstation will be fixed; after changing the operator, if the task and the operator do not conflict, the operator of the workstation will be fixed.

[0040] S3. Perform crossover, mutation and cloning operations on the initial population, update the better solution based on the simulated annealing algorithm and combined with Pareto optimality, until the current temperature of the simulated annealing algorithm is less than the threshold, modify the cooling method of the simulated annealing algorithm to linear cooling and repeat the above steps until the current temperature is less than the final temperature, and output the optimal solution; the linear cooling method is as follows: T1=Tt, T≤t0, where T1 represents the temperature after cooling, T represents the temperature before cooling, t is a fixed cooling value, and t0 is a threshold.

[0041] In this process, crossover mutation is a conventional operation in this field. For example, for the mutation process, single-point mutation is usually selected. The crossover operation is as follows: take the mutated results X1 and X2 as the parent generation, randomly select a part of the sequence on the import side or export side of X1, map and arrange it according to X2, and then put it back to X1 to obtain the offspring of X1. These are all conventional means in this field. And crossover mutation is targeted at the task sequence.

[0042] Since the double-layer encoding method is used in this embodiment, not only the tasks are encoded, but also the operators are encoded; and some tasks have certain requirements for operators. Therefore, in order to expand the search range, in this embodiment, the operator sequence is cloned. When the cloning operation is performed, the task sequence is kept unchanged, and the roulette method is used to select the front workstation as the reverse operator point with a higher probability to generate a new operator sequence, and then the task sequence is assigned to generate a new individual.

[0043] After performing crossover mutation and cloning operations, we use the simulated annealing algorithm to solve the above model.

[0044] Different from the conventional simulated annealing algorithm, the simulated annealing algorithm used in this embodiment is an improved algorithm. The conventional simulated annealing algorithm uses an exponential cooling method, which can be well applied when the temperature is high, but when the temperature reaches a certain threshold, the cooling rate decreases, resulting in a longer convergence time of the algorithm.

[0045] In order to solve this problem, the improved algorithm of this embodiment is a two-stage cooling method: when the current temperature is greater than the threshold, the cooling method is conventional exponential cooling; when the current temperature is lower than the threshold, the cooling method is modified to linear cooling, which improves the convergence speed of the algorithm. For linear cooling, its threshold t0 and fixed cooling value t need to be repeatedly tested according to actual conditions to ensure that the convergence speed of the algorithm is accelerated while the accuracy remains unchanged.

[0046] As for the rest of the operations of the improved algorithm, they are the same as those of the conventional simulated annealing algorithm, such as the setting of the initial temperature and the final temperature, the chain length, etc. At the same time, in order to improve the efficiency of the improved algorithm of the embodiment of the present invention, the present embodiment also uses the Pareto optimality and crowding distance mechanism to update the solution set, and its operation is the same as the conventional operation, so it will not be described in detail in this embodiment.

[0047] To verify the correctness of the model proposed in this study, the operating environment is: Windows 11 system, AMD Ryzen76800H CPU, 3.20GHz, 16GB memory, and Gurobi10.0.2 exact solver is used to calculate the general small-scale example P8 and the general medium-scale example P25. The specific disassembly information is shown in Tables 1 and 2. To verify the rationality and effectiveness of the algorithm constructed in this paper, MATLAB 2021a is used to solve the above examples and compare them with the exact solver. Through a large number of calculations, a set of parameters with better algorithm calculation performance is selected as experimental parameters. The algorithm parameters of the IHCGSAA algorithm for solving small-scale examples are set as follows: initial temperature t_cs=10, final temperature t_zm=1, cooling coefficient q=0.9, linear cooling initial temperature t0=5, population size pop_size=30, chain length l=10, clone number cl=10, and cycle time CT=40s. The algorithm parameters for solving the medium scale are: t_cs = 20, t_zm = 1, q = 0.95, t0 = 5, l = 15, cl = 10, pop_size = 100, and the cycle time CT = 86s. The disassembly information of the P8 example is shown in Table 1, and the disassembly information of the P25 example is shown in Table 2. The purchase cost of the robot is 65,000 yuan. Assuming that the number of products disassembled per day when the disassembly line is running is 160, the estimated operation time of the production line is 50 days. Considering that the power consumption of the robot workstation in the standby state is relatively large, the standby costs of the workers and robots in the two examples are 0.003 yuan and 0.005 yuan per unit time respectively.

[0048] Table 1 P8 example disassembly information

[0049]

[0050] Table 2 P25 example disassembly information

[0051]

[0052] In Table 1 and Table 2, “~” indicates that there is no data for this item.

[0053] The exact solution and algorithm calculation results and solution time of Gurobi are shown in Table 3. The objective functions in the table are: minimizing the number of workstations f1, minimizing the idle time index f2, and minimizing the cost f3. The exact solver and algorithm solution time are the average of the calculation time of 10 independent experiments. The optimal allocation plan for the three objectives of the P25 problem is as follows Figure 1 shown.

[0054] Table 3 Comparison of solution results between Gurobi solver and algorithm

[0055]

[0056] From the comparison in Table 3, it can be seen that for small and medium-sized problems, both the model and the algorithm obtain the same optimal solution for each function within a reasonable time, which effectively proves the correctness of the model in this paper and the effectiveness of the algorithm. By comparing the solution time, it can be seen that with the increase in the scale of the example, the time for the exact solver to solve the model increases significantly. This is due to the NP-hard property of the DLBP problem. At the same time, since the idle time balance index is a nonlinear objective function, the exact solver is difficult to solve, resulting in a large difference in the solution time between the objectives. Compared with the exact solver that can only solve the optimal solution of a single objective, the algorithm of the embodiment of the present invention can solve multiple objective functions at the same time, and can solve the Pareto frontier solution of non-single objective optimal solution, which provides more options for the design of the disassembly line. It can be concluded from the task allocation diagram that the use of the U-shaped disassembly line not only reduces the floor space, but also has a better idle time index when each objective obtains the optimal solution.

[0057] In order to further verify the correctness and superiority of the method of the embodiment of the present invention, and considering that the disassembly objects of the U-shaped disassembly line are mostly small and medium-sized EOL products, the algorithm is applied to a printer disassembly line with hazardous attribute tasks after investigation. The disassembly line has 55 disassembly tasks (P55), and the attributes and priority relationships of each task are shown in Table 4. The simulated disassembly line works 200 days a year, disassembling 80 units per day, and different algorithms are used to calculate and solve it.

[0058] Table 4P55 Disassembly Information Table

[0059]

[0060]

[0061] Hypervolume (HV) is a commonly used comprehensive evaluation index to evaluate the convergence of the Pareto solution set obtained by the multi-objective intelligent algorithm. The higher the HV value, the better the convergence and solution quality of the solution. The calculation formula of the HV value is shown in formula (23):

[0062]

[0063] Where V(F(Xi),G*) represents the hypervolume formed by the solution set and the reference point G*. The reference point is determined by the worst value of each sub-objective of different problems. In this example, the reference point [10, 5000, 1000000] is selected for calculation.

[0064] For large-scale problems, due to the NP-hard property of DLBP, the exact solver cannot obtain the optimal solution within the effective time. Therefore, in order to evaluate the optimization ability and convergence of the algorithm (IHCSSA) of the embodiment of the present invention, it is compared with the classic algorithms: multi-objective genetic algorithm (NSGA-II), simulated annealing algorithm (SA), and immune algorithm (IA). Each algorithm is run ten times, 500 generations each time, and the HV value of each generation is recorded. The HV iteration diagram is drawn as shown in Figure 2 As shown. As can be seen from the figure, in terms of algorithm convergence speed: the algorithm of the embodiment of the present invention begins to converge significantly after 15 iterations. When the algorithm iterates to the 59th generation, the HV index is 6.211×109. After that, the algorithm HV value basically reaches stability, and finally reaches the highest value at the 210th iteration, with an HV index of 6.225×109; IA also has a faster convergence speed. The algorithm basically converges after the 43rd iteration, but its final HV value is 6.133×109, the lowest among the four algorithms; SA's final HV value is 6.213×109, second only to the algorithm of the embodiment of the present invention, but the convergence speed of the SA algorithm is low. The algorithm has not fully converged when it iterates to the 125th generation; the performance of the NSGA-II algorithm is relatively moderate. From the above analysis, it can be seen that the algorithm of the embodiment of the present invention combines the convergence speed of the IA algorithm and the global search performance of the SA algorithm, and has better search performance.

[0065] The algorithm of the embodiment of the present invention is used to solve the example, and a total of 131 Pareto frontier solutions are solved. As shown in Table 5, there are 5 solutions in the Pareto frontier solution, where R in the disassembly scheme indicates that the workstation is a robot workstation, H indicates that the workstation is a human workstation, when the task number is positive, it indicates that the task is located on the import side of the workstation, and when the task number is negative, it indicates that the task is located on the export side of the workstation. According to the analysis of Table 5, the solution range of the target f1 in the solution of the P55 problem is 7 and 8; the solution range of the target f2 is 1 to 851; the solution range of the target f3 is 554296 to 611848. It can be found from the calculation results that the idle time balance index of the U-shaped disassembly line for large-scale instances is still better, with a minimum of 1. When the cost target f3 is optimal, the idle time balance index is 318. At the same time, it is shown that the human-machine collaborative disassembly line mode applied to the U-shaped disassembly line can give full play to the characteristics of the working time difference between different human and machine operators, so that the utilization rate of the cycle time is maximized.

[0066] Table 5 P55 Case Solution Partial Results

[0067]

[0068] Aiming at the problem that the traditional disassembly line does not consider the problem of insufficient constraints for human-machine collaboration, this paper proposes a human-machine collaborative U-shaped disassembly line with multiple constraints, considers the differences in task attributes and working time between human operators and robot operators, and considers AND / OR priority relationship. A multi-objective mathematical model is established with the minimum number of workstations, idle time balance index and long-term cost as the objective function. The Gurobi software is used to accurately solve small and medium-scale examples to verify the correctness of the model. Aiming at the problem in this paper, a double-layer encoding considering the differences between workstation operators is established and a decoding method suitable for this problem is designed. Operations such as clonal mutation in the immune algorithm are introduced into the simulated annealing algorithm, and a two-stage annealing operation is designed to construct an improved hybrid clonal simulated annealing algorithm. According to the characteristics of the U-shaped disassembly line and AND / OR priority relationship, different crossover mutation methods in the global search stage and the local search stage are designed to increase the search ability of the algorithm. By comparing with the results of the exact solver, the correctness and efficiency of the algorithm are verified; by comparing the HV index and IGD index of the solution results with the traditional algorithm, the superiority of the IHCSSA algorithm proposed in this paper is verified.

[0069] The above description is only a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any technician familiar with the profession can make some changes or modifications to equivalent embodiments of equivalent changes using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of the technical solution of the present invention.

Claims

1. A multi-constraint human-machine collaborative U-shaped disassembly line optimization method, characterized in that: The following steps are involved: S1. Establish a disassembly model, wherein the disassembly model includes an objective function and constraints. The objective function is to minimize the number of open workstations, the idle time of the disassembly line, and the disassembly line cost. The disassembly line cost includes the machine purchase cost and the disassembly line operation cost. The constraints include task allocation constraints, workstation and disassembler allocation constraints, complex task and hazardous task allocation constraints, takt time constraints, predecessor task constraints, and workstation opening constraints. S2, generating an initial population based on the disassembly task and constraints; S3. Perform crossover, mutation and cloning operations on the initial population, update the better solution based on the simulated annealing algorithm and combined with Pareto optimality, until the current temperature of the simulated annealing algorithm is less than the threshold, modify the cooling method of the simulated annealing algorithm to linear cooling and repeat the above steps until the current temperature is less than the final temperature, and output the optimal solution; the linear cooling method is as follows: T1=Tt, T≤t0, where T1 represents the temperature after cooling, T represents the temperature before cooling, t is a fixed cooling value, and t0 is a threshold.

2. The method according to claim 1, characterized in that The objective function is as follows: Where F represents the minimization objective; f1 represents the number of open workstations; f2 represents the idle time of the disassembly line; f3 represents the cost of the disassembly line; N is the set of workstations, and n∈N; l represents the operator, when l=1, it means the operator is a worker, and when l=2, it means the operator is a machine; z ln is a binary variable, z ln =1 means the nth workstation is open and the operator is l, otherwise z ln =0; CT represents the cycle time; m represents the import and export sides of the U-shaped disassembly line, where m=1 represents the import side and m=2 represents the export side; represents the disassembly time of task i when the operator is l; is a binary variable, Indicates that task i is assigned to port m of the nth worker station of operator l, otherwise C r represents the purchase cost of a unit machine; D represents the number of days the U-shaped disassembly line is in operation; G represents the total number of parts disassembled per day; C d represents the cost of disassembling each product; represents the cost of disassembling task i for operator l; It represents the standby cost of the workstation when the operator is l.

3. The method according to claim 1, characterized in that The constraints include: Task allocation constraints: Where N is the set of workstations, and n∈N; m represents the import and export sides of the U-shaped disassembly line, where m=1 represents the import side and m=2 represents the export side; l represents the operator, when l=1, it means the operator is a worker, and when l=2, it means the operator is a machine; is a binary variable, Indicates that task i is assigned to port m of the nth worker station of operator l, otherwise Workstation and disassembler allocation constraints: In the formula, M is greater than, z ln is a binary variable, z ln =1 means the nth workstation is open and the operator is l, otherwise z ln =0; Complex and hazardous task allocation constraints: In the formula, K represents the complex task set; H represents the hazardous task set; Takt time constraints: In the formula, CT represents the takt time; Predecessor task constraints: Where TP ij Represents the task priority relationship. If task i is the AND immediate predecessor of task j, then TP ij =1, if task i is the OR immediate predecessor of task j, then TP ij =-1, otherwise TP ij =0;OR j represents the set of OR immediate predecessor tasks of task j; y ij is a binary variable, y ij =1 indicates that task i is the OR immediate predecessor of task j and task i is completed before task j, otherwise y ij =0; is a binary variable, indicates that task i is assigned to the import of the nth worker station of operator l, otherwise is a binary variable, indicates that task j is assigned to the import of the nth worker station of operator l, otherwise is a binary variable, Indicates that task i is assigned to the exit of the nth worker station of operator l, otherwise is a binary variable, indicates that task j is assigned to the exit of the nth worker station of operator l, otherwise Workstation startup constraints:

4. The method according to claim 1, characterized in that: In S2, generating the initial population includes encoding and decoding steps; The encoding step includes the following sub-steps: Based on the constraints, firstly, the tasks without immediate predecessors are assigned to the entrance of the disassembly line, and their task sequences are marked as positive numbers, and the tasks without immediate future tasks are assigned to the exit of the disassembly line, and their task sequences are marked as negative numbers; Then update the priority relationship of the remaining tasks, and repeat the above operation until all tasks are assigned; Combined with the operator sequence of the workstation, a two-layer code is generated; The decoding step comprises the following sub-steps: Assign operators to workstations according to the operator sequence and assign tasks until the takt time is reached; Check whether there is a conflict between the task and the operator. If there is a conflict, change the operator; After changing the operator, if the non-primary task and the operator still conflict, the operator will be changed back to the original operator; after changing the operator, if the primary task and the operator conflict, the operator of the workstation will be fixed; after changing the operator, if the task and the operator do not conflict, the operator of the workstation will be fixed.

5. The method according to claim 1, characterized in that In S3, the operator sequence is cloned. When the cloning operation is performed, a roulette wheel method is used to select the front workstation as the reversal operator point with a higher probability.

6. The method according to claim 1, characterized in that In the exponential cooling stage of the simulated annealing algorithm, the mutation operation is a single-point mutation, and the crossover operation is as follows: take the mutated results X1 and X2 as the parent generation, randomly select a part of the sequence on the import side or export side of X1, map and arrange it according to X2, and then put it back to X1 to obtain the offspring of X1.

7. The method according to claim 1, characterized in that In the linear cooling stage of the simulated annealing algorithm, the mutation operation is a three-point mutation, and the crossover operation is as follows: take the mutated results X1 and X2 as parents, take the fragment in X1, put the tasks in the fragment that are the same as the import and export of X2 back into the sequence X1, and randomly insert the remaining tasks according to the constraints to obtain the offspring of X1.

Images