A human-computer interaction constrained bilateral disassembly line setting method

Through the hybrid integer linear programming model and discrete squirrel search algorithm, the two-sided disassembly line settings of human-computer interaction constraints are solved, and the problem that human-computer collaboration factors in the existing technology are not fully considered, which improves disassembly efficiency and reduces costs, and achieves more Excellent disassembly line combination.

CN116702624BActive Publication Date: 2024-11-22SOUTHWEST JIAOTONG UNIV
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202310762510.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2023-04-07
Filing Date
2023-06-27
Publication Date
2024-11-22
Estimated Expiration
2043-06-27

AI Technical Summary

Technical Problem

The prior art fails to fully consider the human-computer collaboration factors in the disassembly line balance problem in human-computer interaction technology, resulting in low disassembly efficiency, high cost, and difficulty in optimizing the setting of bilateral disassembly lines.

Method used

A method for setting the bilateral disassembly line of human-computer interaction constraints is proposed. Through a mixed integer linear planning model and discrete squirrel search algorithm, the number of workstations, idle time balance, robot energy consumption and disassembly cost are optimized, and combined with human-computer interaction constraint operations, a A disassembly line combination that is more suitable for actual work scenarios.

Benefits of technology

It realizes the single-objective optimal solution for small-scale cases quickly, and through fast non-dominant sorting and discrete squirrel search algorithm, the solution ability of large-scale cases is improved, and the better disassembly line combination is obtained, and the metaheuristic algorithm is verified. correctness.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116702624B_ABST
    Figure CN116702624B_ABST
Patent Text Reader

Abstract

A method for setting up bilateral disassembly lines with human-machine interaction constraints comprises the following steps: collecting disassembly line and disassembly task information, establishing a mixed integer linear programming mathematical model for the bilateral disassembly line balancing problem with the goal of minimizing the number of open workstations, idle time balance indicators, robot energy consumption and labor costs, and including human-machine interaction constraint operations; then sorting and classifying the population through fast non-dominant sorting, optimizing the classified initial solution population using a discrete squirrel search algorithm, and outputting a final solution that meets the iteration conditions as a disassembly solution for execution; the present invention introduces a human-machine interaction disassembly mode into the disassembly line balancing problem for the first time, and establishes a MIP model that considers human-machine interaction behavior, which can solve the single-objective optimal solution of a small-scale case in a very short time, and can also verify the correctness of the proposed metaheuristic algorithm; the proposed discrete squirrel search algorithm can quickly obtain a true Pareto frontier solution, especially for solving large-scale cases.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of facility layout, in particular to a method for setting a bilateral disassembly line with human-machine interaction constraints. Background Art

[0002] The decreasing primary resources, strict environmental regulations, and the improvement of environmental awareness have promoted the rapid development of the circular economy worldwide. As the main carrier of the circular economy, the remanufacturing industry will face more challenges in many aspects such as the transportation, disassembly and recycling of renewable resources, and the green supply chain. Disassembly and classification is the most critical link in the remanufacturing process, because scrapped products can be recycled, reused or remanufactured after being dismantled, which effectively extends the life cycle of scrapped products and creates economic value. Due to the advantages of standardization and high efficiency in disassembling large-scale end-of-life (EOL) products, disassembly lines have been widely accepted and adopted by resource recycling companies.

[0003] Large-scale EOL products are usually large in size, and single-sided or U-shaped disassembly lines may reduce the disassembly efficiency of operators. For example, when operators disassemble the tires and doors of a car, they need to move back and forth on both sides of the disassembly line, which increases invalid production time. However, a double-sided disassembly line can set up pairing stations on both sides. The pairing stations can share operating tools and fixtures and complete the disassembly tasks on both sides at the same time. This layout not only shortens the line length but also reduces production time. In addition, when disassembling large-scale products, some disassembly tasks require the collaboration of operators on both sides. For example, when disassembling the windshield of a car, operators on both sides are first required to peel off the surrounding sealing rubber rings at the same time, and then lift the windshield off the workstation collaboratively. According to the survey of existing literature, past studies usually assume that a task is assigned to only one operator, and do not consider the same task being completed by two operators in collaboration, although collaborative disassembly often occurs in practice.

[0004] Traditional disassembly methods usually assume that disassembly tasks are completed by humans, and consider the differences between different operators when establishing mathematical models, including physical strength, health status, skill level, etc. Nowadays, with the rapid iteration and upgrading of robot technology, some robots can replace humans to complete simple and repetitive disassembly tasks. With the continuous wave of Industry 5.0, Industry 5.0 requires a more personalized human-machine interaction technology to complement the advantages of humans and machines and protect people's physical and mental health and well-being in production and manufacturing. However, human-machine interaction technology adds many constraints to the disassembly line balancing problem (DLBP), such as the constraint relationship between humans and machines, and the constraint relationship between humans, machines and products. These constraints dramatically increase the difficulty and complexity of DLBP. In solving the problem of disassembly line balancing involving human-machine interaction technology, it is necessary to consider relevant factors more comprehensively. Summary of the invention

[0005] In view of this, the main purpose of the present invention is to provide a human-machine interaction constrained bilateral disassembly line setting method (HRI-TDLBP), which takes into account the influence of the number of workstations, idle balance indicators, robot energy consumption and disassembly cost on the disassembly line setting, is more in line with the actual working conditions of human-machine disassembly, and can obtain a better disassembly line combination.

[0006] The technical solution of the present invention is a method for setting a bilateral disassembly line with human-machine interaction constraints, comprising the following steps:

[0007] Step S1: Collect disassembly line and disassembly task information, and establish a mixed integer linear programming mathematical model for the bilateral disassembly line balancing problem with the goal of minimizing the number of open workstations, idle time balance indicators, robot energy consumption and labor costs, including human-machine interaction constraint operations.

[0008] Step S2: Initialize the population, then sort and classify the population through fast non-dominant sorting, and establish an external archive to record the initial solution population after classification;

[0009] Step S3: The classified initial solution population is optimized using the discrete squirrel search algorithm, and the final solution that meets the iteration conditions is output as the disassembly plan for execution.

[0010] The technical effects of the present invention are:

[0011] 1. This paper introduces the human-computer interactive disassembly mode into the disassembly line balance problem for the first time, and establishes a mixed integer programming model (Mixed Integer Program, MIP) considering the human-computer interactive behavior, so that the single-objective optimal solution of small-scale cases can be solved in a very short time, and is also used to verify the correctness of the proposed metaheuristic algorithm.

[0012] 2. The present invention designs encoding and decoding according to the characteristics of the problem, and introduces fast non-dominated sorting to improve the search ability of the algorithm. The innovative discrete squirrel search algorithm (DSSA) can quickly obtain the true Pareto front solution, especially for solving large-scale cases. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below.

[0014] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0015] Figure 2 It is a schematic diagram of a bilateral disassembly line with human-machine interaction behavior constraints of the present invention;

[0016] Figure 3A schematic diagram of converting the priority relationship diagram in the present invention into a priority relationship matrix;

[0017] Figure 4 A schematic diagram of the process of performing an exchange task between two sequences in DSSA through a subtraction operation in the present invention;

[0018] Figure 5 Find the Pareto frontier diagrams for solving P52 for different algorithms;

[0019] Figure 6 This is a three-dimensional image of a certain brand of power battery module;

[0020] Figure 7 The solution set distribution of different algorithms for the power battery module case;

[0021] Figure 8 Four disassembly plans for the power battery module. DETAILED DESCRIPTION

[0022] The present invention will be further described in detail below in conjunction with the embodiments and drawings.

[0023] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] Example:

[0025] See also Figure 1 , Figure 2 The specific process of the human-machine interaction constraint bilateral disassembly line setting method proposed in the present invention is as follows: Figure 1 As shown, the following steps are included:

[0026] Step S1: Collect disassembly line and disassembly task information, and establish a mixed integer linear programming mathematical model for the bilateral disassembly line balancing problem with the goal of minimizing the number of open workstations, idle time balance indicators, robot energy consumption and labor costs, including human-machine interaction constraint operations.

[0027] The mathematical model includes the disassembly time, part attributes and disassembly priority relationship of any part in the disassembled product, wherein the disassembly time includes the disassembly time when disassembly is performed by workers and the disassembly time when disassembly is performed by robots.

[0028] Part attributes include hazardous parts, complex parts and ordinary parts. Hazardous parts refer to parts that have certain safety hazards during the disassembly process and need to be assigned to robots for disassembly as much as possible; complex parts refer to parts with complex structures that are difficult for robots to disassemble and usually need to be assigned to workers for disassembly; ordinary parts refer to more conventional parts and can be assigned to robots or workers for disassembly.

[0029] The disassembly priority relationship means that for some parts, other parts must be disassembled before the part can be disassembled. Therefore, the other parts are called the immediate predecessor tasks of the part.

[0030] It can be seen that the objective function includes minimizing the number of open workstations, idle time balance index, robot workstation energy consumption and disassembly cost. Its objective function should be:

[0031] F=min(f1,f2,f3,f4) (1)

[0032] The minimum number of open workstations is shown in formula (2):

[0033]

[0034] The minimum idle time balance index is shown in formula (3):

[0035]

[0036] After linearizing equation (3), the minimum idle time balance index after linearization constraint is obtained as shown in equation (4):

[0037]

[0038] Where Q = |y iw -y jw |

[0039] Minimizing the energy consumption of the robot workstation is shown in formula (5):

[0040]

[0041] After linearizing equation (5), the energy consumption of the robot workstation minimized after linearization constraints is obtained as shown in equation (6):

[0042]

[0043] in,

[0044] The minimum disassembly cost is shown in formula (7):

[0045]

[0046] In the above formulas, S k is the workstation enable variable, indicating that if the kth workstation is enabled, then S k =1, otherwise S k =0; CT is the cycle time; I is the disassembly task set, {1,2,...,n}; M is the pairing workstation number set; L is the pairing station number, L={l|1,2}, l=1 means the left side, l=2 means the right side; W is the operator number, W={w|1,2}, w=1 means worker disassembly, w=2 means robot disassembly; i, j means disassembly task number; is a binary variable, indicating that if task i is assigned to the l-side of the k-th pairing station, then otherwise t i refers to the disassembly time of task i; OE refers to the operating power consumption of the robot per unit time; refers to the operator assignment variable. If task i is assigned to w operators at the kth pairing station, then otherwise, Where y i1 Assign variables to the operator when the operator is a worker, y i2 The variable is assigned to the operator when the operator is a robot; SE refers to the robot's standby power consumption per unit time; SR kl Assign variables to the robot. If the l-side of the k-th paired station is a robot, then SR kl =1, otherwise, SR kl =0; C rt The unit time cost of robot disassembly; C mt is the unit time cost of workers when disassembling; BH is the set of tasks with hazardous attributes in the product to be disassembled; C h It refers to the additional cost required for dismantling hazardous tasks;

[0047] The first goal f1 in the mathematical model of the disassembly line is given by equation (2), which means minimizing the number of workstations to reduce costs. The second goal f2 is given by equation (3), which aims to improve the load smoothness between workstations and reduce the idle time of workstations to increase disassembly efficiency. The third goal f3 is given by equation (5), which aims to minimize the energy consumption of the robot workstation, including the standby energy consumption and disassembly energy consumption of the robot. The fourth goal f4 is given by equation (7), which is used to measure the disassembly cost, including the robot disassembly cost, the worker disassembly cost and the additional cost of the disassembly hazard task.

[0048] In order to improve the solution efficiency of the MIP model, it is necessary to linearize the model, where:

[0049] The objective f2 represented in equation (3) can be added with the linearization constraints required to be satisfied for linearization processing of the minimization idle time balancing index as shown in equations (8)-(13):

[0050] Q ijw =max{y jw -y iw ,y iw -y jw} (8)

[0051]

[0052]

[0053]

[0054]

[0055]

[0056] The target f3 represented by formula (5) contains and iw The product of two decision variables, let Then the target f3 can be expressed as shown in equation (6), and the corresponding linear constraints are added, such as the linearization constraints that need to be satisfied for the linearization processing to minimize the energy consumption of the robot workstation as shown in equations (14)-(17):

[0057]

[0058]

[0059]

[0060]

[0061] Among them, P iklw Represents the variables required for linearization of the objective function.

[0062] In the present invention, the objective function also needs to satisfy corresponding constraints, which are specifically basic constraints, direction constraints, task attribute constraints and human-computer interaction constraints.

[0063] Basic constraints are constraints that all types of disassembly lines need to meet, which means that when the product flow enters the disassembly line and is decomposed, the parts priority relationship and cycle time constraints must be met. In a bilateral disassembly line, the disassembly tasks are assigned to the pairing stations on the left and right sides. The tasks between the two pairing stations must meet the priority relationship, and the sum of the completion time of all tasks of a single pairing station must be less than the cycle time.

[0064] Direction constraints are determined by the characteristics of the bilateral disassembly line, which means that the task set to be assigned is divided into three directions: the left task set, the right task set, and other task sets. The left task can only be assigned to the left side of the line, the right task set can only be assigned to the right side of the line, and other tasks can be assigned to any side.

[0065] The task attribute constraints refer to the safety strategy considered in the human-machine hybrid disassembly mode. The present invention stipulates that hazardous tasks must be assigned to robots, while complex tasks must be assigned to humans. Tasks that are both hazardous and complex are defined as complex tasks because robots cannot complete them.

[0066] The interaction behavior in human-computer interaction constraints can be defined as follows:

[0067] (1) Human-human interaction: When the operators on both sides are determined to be manual, the interactive disassembly is defined as human-human interaction behavior, and the interactive task can only be a complex task or a task without attributes.

[0068] (2) Human-machine interaction: When the operator on one side is human and the operator on the other side is a robot, the interactive disassembly is defined as a human-machine interactive behavior, and the interactive task can only be a non-attribute task.

[0069] (3) Machine-machine interaction: When the operators on both sides are determined to be robots, the interactive disassembly is defined as a machine-machine interaction behavior, and the interactive tasks can only be hazardous tasks or non-attribute tasks.

[0070] Based on this, the above constraints should satisfy the following formula:

[0071] Task allocation constraints:

[0072]

[0073]

[0074]

[0075]

[0076]

[0077] Formulas (18)-(22) represent the task allocation constraints. Formula (18) assigns all tasks to workstations. Formulas (19) and (20) restrict the left-hand and right-hand tasks to be assigned to the left and right sides of the workstation, respectively. Formula (21) indicates that the interactive disassembly task must be assigned to the paired stations on both sides of the same workstation. Formula (22) indicates that complex tasks and hazardous tasks cannot be assigned to the same paired station.

[0078] Operation allocation constraints:

[0079]

[0080]

[0081]

[0082]

[0083] Formulas (23)-(26) represent task allocation constraints. Formula (23) restricts a task to be assigned to only one operator. Formulas (24) and (25) require that complex tasks must be assigned to humans, while hazardous tasks must be assigned to robots. Formula (26) indicates that tasks in the same pairing station can only be completed by one operator.

[0084] Time constraints:

[0085]

[0086]

[0087]

[0088]

[0089]

[0090]

[0091]

[0092] Formulas (27)-(33) represent time constraints. Formula (27) provides the upper and lower bounds of the number of workstations based on the task time, cycle time, and total number of tasks. Formulas (28) and (29) indicate that the tasks of the same workstation must be completed in sequence. Formula (30) limits the working time of each paired station to not exceed the cycle time, and formula (31) indicates that the start time of the interactive tasks must be consistent. Formula (32) limits the start time of the subsequent task to be later than the end time of the preceding task. Formula (33) gives a lower bound on the task start time.

[0093] Precedence constraint:

[0094]

[0095]

[0096] Workstation allocation constraints:

[0097]

[0098]

[0099]

[0100]

[0101]

[0102] Formulas (36)-(40) represent the workstation allocation constraints. Formula (36) indicates that if the current paired station has a hazardous task, the operator is a robot. Similarly, Formula (37) indicates that if the current paired station has a complex task, the operator is a human. Formula (38) indicates that only one operator can be assigned to the same paired station. Formula (39) constrains the upper and lower bounds of the number of workstations. Formula (40) indicates that the workstations must be opened in sequence.

[0103] Binary variable range:

[0104]

[0105] In the above formulas, S k is the workstation enable variable, indicating that if the kth workstation is enabled, then S k =1, otherwise S k =0; CT is the cycle time; I is the disassembly task set, {1,2,...,n}; M is the pairing workstation number set; L is the pairing station number, L={l|1,2}, l=1 means the left side, l=2 means the right side; W is the operator number, W={w|1,2}, w=1 means worker disassembly, w=2 means robot disassembly; i, j means disassembly task number; is a binary variable, indicating that if task i is assigned to the l-side of the k-th pairing station, then otherwise t i refers to the disassembly time of task i; OE refers to the operating power consumption of the robot per unit time; refers to the operator assignment variable. If task i is assigned to w operators at the kth pairing station, then otherwise, Where y i1 Assign variables to the operator when the operator is a worker, y i2 The variable is assigned to the operator when the operator is a robot; SE refers to the robot's standby power consumption per unit time; SR kl Assign variables to the robot. If the l-side of the k-th paired station is a robot, then SR kl =1, otherwise, SR kl =0; C rt The unit time cost of robot disassembly; C mtThe unit time cost of workers when disassembling; C h Refers to the additional cost required to disassemble hazardous tasks; BI refers to the set of tasks that require interactive disassembly by workstations on both sides of the paired station in the product to be disassembled; BC refers to the set of tasks with complex attributes in the product to be disassembled; BH refers to the set of tasks with hazardous attributes in the product to be disassembled; i', j' are the numbers of the tasks that require system disassembly; T l Assign a set of tasks to each side of the paired station, if task i∈T l , then i can only be assigned to the l side of the paired station; is a positive number; ST i refers to the start time of task i, ST j Similarly; TP ij Refers to the priority relationship matrix of the disassembly task, TP ij =[a ij ] n×n ; if a ij =1, task i is the immediate predecessor of task j; is a disassembly sequence variable. If tasks i and j are assigned to w operators located on the l side of the k-th pairing station and task i is assigned before task j, then otherwise, Q ijw and u ijw All are variables required for constrained linearization;

[0106] Step S2: Initialize the population, then sort and classify the population through fast non-dominant sorting, and establish an external archive to record the initial solution population after classification;

[0107] The population initialization process includes encoding, decoding and objective function calculation. The encoding method used in the present invention can be used as follows: Figure 3 A 10-scale case is shown for illustration. Figure 3 This is the process of converting a 10-scale task priority relationship diagram into a priority relationship matrix, in which Task 7 can be disassembled only after Tasks 5 and 6 are disassembled, so Tasks 5 and 6 are called the immediate predecessor tasks of Task 7; at the same time, Tasks 2 and 3 can be disassembled only after Task 8 is disassembled, so Tasks 2 and 3 are considered to be the immediate successor tasks of Task 8.

[0108] Obtain the initial solution based on the TP matrix: First, find tasks without any immediate predecessor constraints, that is, all elements of the column vector corresponding to the task number in the TP matrix are 0, and randomly sort these tasks. When i is assigned, set all elements in the i-th row of the TP matrix to zero to remove the restriction of task i on the immediately following task j. Finally, repeat the above operation to obtain a set of feasible disassembly sequences with an array length equal to the task size.

[0109] After obtaining a feasible disassembly sequence, the corresponding disassembly tasks are first identified by their attributes. According to the model constraints, tasks with complex attributes and hazardous attributes must be assigned to manual workstations and robot workstations respectively, that is, the workstation type is determined by the task attributes. If a workstation is first assigned a task without any attributes, the workstation type is not defined temporarily until a task with hazardous or complex attributes is assigned, and then the workstation type is defined according to the task attributes. When the workstation type is determined, new workstations must be opened for tasks with different attributes.

[0110] After that, an external archive WD is established and updated, and the individuals in the generated population are sorted quickly and non-dominated based on the Pareto elite strategy. The first type of solution, that is, the optimal solution, is screened out through the Pareto elite strategy; the remaining solutions are screened by the Pareto elite strategy to obtain the second type of non-inferior solutions; the remaining poor solutions are taken as the third type of solutions, and three types of populations are graded for the optimization process of the discrete squirrel search algorithm (DSSA) in the subsequent steps.

[0111] Step S3: The classified initial solution population is optimized using the discrete squirrel search algorithm, and the final solution that meets the iteration conditions is output as the disassembly plan.

[0112] The Squirrel Search Algorithm (SSA) is a currently commonly used optimization algorithm with good experimental results. In order to adapt to the application scenario in the present invention, a discrete adjustment is made to the SSA.

[0113] The discrete squirrel search algorithm (DSSA) with discreteness adjustment mainly includes the following steps:

[0114] 1. First, set the problem parameters and algorithm parameters. The problem parameters include the preset cycle time CT, the energy consumption of the robot workstation (including the standby energy consumption and disassembly energy consumption of the robot), and the disassembly cost (including the robot disassembly cost, the worker disassembly cost, and the additional cost of the disassembly hazard task); the algorithm parameters include the population size Pop_num, the maximum number of iterations G, and the number of external archives N;

[0115] 2. Construct the initial population of the task disassembly sequence after encoding in step S2, create an external archive, and start iteration;

[0116] 3. Then, the position is updated by randomly initializing it through non-dominant sorting to achieve different operation choices;

[0117] like Figure 1As shown in the figure, the population obtained after initialization is classified by non-dominant sorting to achieve three corresponding operation behaviors. The three specific operation behaviors are analogous to the foraging behavior of flying squirrels: in the forest, there are only three kinds of trees for flying squirrels to choose from, such as ordinary trees (no food), oak trees (source of acorn nuts) and hickory trees (source of hickory nuts). One of the three specific operation behaviors is selected by random numbers to achieve position update, and a partial random solution is generated at the same time, as follows:

[0118] According to the SSA algorithm, the migratory behavior of flying squirrels is always affected by the presence of predators. When there are no predators, the flying squirrel will effectively glide and search for its favorite food throughout the forest. When predators are present, the flying squirrel is forced to use small random walks to search for nearby hiding places. The probability of predator presence P can be set dp The position update mechanism is used to simulate this natural behavior. g is the random sliding distance, R1, R2 and R3 are random numbers obtained in the interval [0,1], FS ht It is the location of the flying squirrel that arrived at the hickory tree, FS at It is the position of the flying squirrel moving from the acorn tree to the hickory tree, FS nt is the position of the flying squirrel on the ordinary tree moving to the acorn tree to meet daily energy needs, t represents the current iteration, G c is the glide constant, which is used to balance exploration and development. The three cases of dynamic migration behavior of flying squirrels in the conventional SSA algorithm are as follows:

[0119] Case 1: The flying squirrel may move towards the hickory tree, and the new position is updated as follows:

[0120]

[0121] Case 2: The flying squirrel moves to the acorn tree to meet its daily energy needs. The new position is updated as follows:

[0122]

[0123] Scenario 3: The flying squirrel has already consumed the acorns and may move to the pecan tree to store pecans in case of food shortage. The new position is updated as follows:

[0124]

[0125] Since the HRI-TDLBP in the present invention is a discrete problem, continuous operation is not applicable, so the three migration behaviors are discretized as follows:

[0126]

[0127]

[0128]

[0129] In formulas (44)-(46), ⊙ represents a feasible exchange pair between two sequences obtained by subtraction operation, It means that the feasible exchange task set is obtained by multiplying the coefficients. Indicates that the current sequence and the exchange task set are exchanged through addition operations. For specific examples, please refer to Figure 4 , sequence FS at With sequence FS ht The exchange pair FS is obtained by subtraction operation at ⊙FS ht ={(9,6),(10,9),(6,10)}. It can be seen that tasks 9, 10, and 6 are in the sequence FS at The commutative intervals in are (1, 4, 5, 10, 6, 7, 8), (1, 4, 9, 5, 6, 7, 8) and (1, 4, 9, 5, 10, 7, 8). Tasks 6, 9, and 10 are in the sequence FS. ht The exchangeable intervals in are (1, 4, 5, 9, 10, 7, 8), (1, 4, 6, 5, 10, 7, 8) and (1, 4, 6, 5, 9, 7, 8), respectively. The exchangeable intervals are determined according to the priority relationship of the disassembly tasks. If both tasks of the exchange pair appear in the exchange interval of the other party, it means that they are feasible. On the contrary, if one task in the exchange pair is not in the exchangeable interval of the other party, then the exchange pair is not feasible. Finally, the new sequence is obtained through the feasible exchange pairs. Similarly, referring to the steps of the above subtraction operation, a corresponding exchange task set after the multiplication operation and the addition operation can also be obtained according to the prior art.

[0130] 4. After the position is updated, seasonal monitoring is also introduced into the DSSA algorithm to prevent the algorithm from falling into the local optimum;

[0131] Collect, summarize and integrate the solutions that achieve position changes in the previous step and the generated random solutions, and calculate the seasonal constants through the formula. By comparing the size of the seasonal constants, decide whether to implement the seasonal perturbation strategy to ensure that the local area can be jumped out again in the final optimization stage. The search group can be randomly bounced to a new area far away from the current location to check whether there is a better target solution. The specific operations of seasonal monitoring are as follows:

[0132] Seasonal changes significantly affect the foraging activities of flying squirrels. Flying squirrels have high foraging costs in low-temperature environments and suffer huge heat losses. Climate conditions force them to be inactive in winter. The seasonal monitoring perturbation strategy simulates this weather influence and prevents SSA from being trapped in a local optimal solution.

[0133] In the present invention, the seasonal constant is calculated based on the discrete characteristics of the problem. The calculation formula of is modified as shown in formula (47):

[0134]

[0135] In the formula, m represents and FS ht The number of exchange pairs, x1 and x2 represent the first and second tasks of the exchange pair, G m It means giving the current task a random number in the interval [0,1]. FS ht,k are fitness values.

[0136] Seasonal monitoring conditions are Where S min is the minimum value of the seasonal constant, and its expression is shown in formula (48):

[0137]

[0138] In formula (48), t and t m are the current iteration and the maximum number of iterations respectively. As we can see, S min Related to the exploration and development capabilities of DSSA. min The larger the value, the more exploratory the algorithm is. min The smaller the value, the stronger the algorithm's exploitation capability.

[0139] 5. After satisfying seasonal monitoring, the Levy flight strategy is introduced to further help the algorithm to better perform global search;

[0140] If it is found that the seasonal monitoring conditions have been met (i.e., winter has ended), the flying squirrels that are unable to find the best winter food source in the forest will migrate arbitrarily. At the end of winter, flying squirrels become active due to the low foraging cost. Flying squirrels that are unable to find the best food source in the forest in winter but still survive may forage in new directions. Incorporating this behavior into the modeling can improve the exploration ability of DSSA. It is assumed that only those flying squirrels that cannot find the pecan kernel food source and survive will move in different directions to find a better food source. The position update of this flying squirrel is shown in Equation (49):

[0141]

[0142] In formula (49), The new location of the flying squirrels moving from the ordinary trees to the acorn trees to meet their daily energy needs; FS L and FS Uare the lower and upper bounds of the position of the i-th flying squirrel, ⊕ indicates that the current sequence is exchanged with the exchange task set through addition operation, and ⊙ indicates that a feasible exchange pair between two sequences is obtained through subtraction operation.

[0143] Here, Levy distribution is used to promote better and more efficient exploration of the solution space. Its mathematical expression is shown in formula (50):

[0144]

[0145] In formula (50), γ, μ> 0. γ is a scale parameter, μ is a shift parameter, and s is a constant close to positive infinity.

[0146] The calculation formula of Levy flight is shown in formula (51):

[0147]

[0148] Where: r a and r b is a normally distributed random number on [0,1]; β is a constant, which is 1.5 in DSSA;

[0149] The calculation formula of σ is shown in formula (52):

[0150]

[0151] Among them, Γ(n)=(n-1)!

[0152] 6. Finally, the optimization process of the discrete squirrel search algorithm (DSSA) is repeated to obtain individuals with better feasibility of the optimal solution, and the external archive WD is updated until the number of iterative calculations meets the termination condition, and the fast non-dominant sorting based on the Pareto elite strategy is used again to obtain the non-inferior solution that meets the iterative condition in the external archive as the final result output.

[0153] The following is a demonstration of the actual operation of this embodiment:

[0154] Program testing environment:

[0155] The simulation computing environment for the numerical experiments in this chapter is Intel(R) Core(TM) i9-10900K CPU@3.7GHz, 32GBRAM Windows 10Pro operating system, and MATLAB R2020b is used for programming and running.

[0156] In order to verify the solving performance of the proposed DSSA, a classic sewing machine case (P52) is introduced, which has 52 tasks. The optimization objectives of the P52 case include the idle rate (F Idle ), smoothing rate (F Smooth ) and disassembly cost (FCost ). The large-scale instance P52 containing 52 sewing machine disassembly tasks is solved using the same type of algorithms: Ant Colony / Genetic Algorithm (ACGA), Genetic Simulated Annealing Algorithm (GASA), Artificial Fish Swarm Algorithm (AFSA), Multi-Objective Improved Particle Swarm Optimization (MIPSO) and Local Neighborhood Genetic Algorithm (LNGA). After a large number of tests, this study sets the DSSA parameter to N S =500,P dp =0.2, the algorithm deadline is 500s, and the results are shown in Table 1.

[0157] Table 1 Results of solving HRI-DLBP with different algorithms

[0158]

[0159] Table 1 lists the results of DSSA and the other five algorithms on the sewing machine case. After a Pareto screening of all the solutions in Table 1, only 14 solutions are retained, of which 3 solutions belong to the solution set of LNGA and 11 solutions belong to the solution set obtained by the proposed DSSA, as shown in bold font in Table 1. Since the first target of all the algorithms' solutions is 0.0579, only targets 2 and 3 are used as coordinate systems to draw the distribution diagram of all algorithms in the target space. Figure 5 As shown, it can be clearly seen that the solution of the proposed DSSA becomes the Pareto frontier of other algorithms.

[0160] In order to further verify the superiority of DSSA, this paper introduces a disassembly case of a certain brand of power battery module to verify the proposed algorithm's ability to solve the model and its practical application. At the same time, it is compared with the flower pollination algorithm (FPA), non-dominated sorting genetic algorithm (NSGA-II), genetic simulated annealing algorithm (GASA) and artificial fish swarm algorithm (AFSA), and the parameter settings of each algorithm are shown in Table 2.

[0161] Table 2 Parameter settings of different algorithms

[0162]

[0163]

[0164] Figure 6 The three-dimensional image of a certain brand of power battery module is shown, which contains 44 disassembly tasks. The names of the components corresponding to the numbers in the figure are shown in Table 3, which gives the disassembly time, disassembly direction, interaction constraints, task attributes and priority relationships of each task.

[0165] Table 3 Part names corresponding to numbers

[0166]

[0167]

[0168] The distribution of Pareto solutions in the target space can intuitively show the situation close to the Pareto frontier. For this reason, the 20 Pareto solutions obtained by each algorithm using crowding distance are placed in the target space. It should be noted that since this study considers four objectives, in order to facilitate analysis, the four objectives are combined into three groups (f1, f2, f3; f1, f2, f4; f1, f3, f4; f2, f3, f4) and then Figure 7 To show the distribution of solution sets of different algorithms in the target space.

[0169] Figure 7 The distribution of the solution sets obtained by DSSA, FPA, NSGA-II, GASA and AFSA in the target space is shown. Figure 7 From (3) and (4), we can see that objectives 3 and 4 are two extremely conflicting objectives, and the solutions obtained by each algorithm are almost linear in the objective space. However, from the four subgraphs, we can intuitively see that compared with FPA, NSGA-II, GASA and AFSA, the solution set obtained by DSSA obviously occupies more Pareto frontier positions.

[0170] Considering that different decision makers have different preferences for goals, the present invention selects four different solutions from the obtained Pareto solution set, such as Figure 8 As shown in the figure, from top to bottom, scheme 1, f1 = 11 f2 = 1691 f3 = 8515.2 f4 = 22358; scheme 2, f1 = 11 f2 = 1691 f3 = 9015.2 f4 = 19158; scheme 3, f1 = 11 f2 = 13901 f3 = 8107.2 f4 = 25078; scheme 4, f1 = 11 f2 = 1847 f3 = 9395.2 f4 = 16758. It can be seen that these schemes all meet the set CT constraints, direction constraints, task attribute constraints and human-computer interaction constraints. The number of workstations and the value of smoothness index of schemes 1 and 2 are both optimal. However, scheme 3 consumes the least energy for the robot, and scheme 4 minimizes the labor cost.

[0171] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the embodiments of the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A human-machine interactive constraint bilateral disassembly line setting method, characterized in that: The following steps are involved: Step S1: Collect disassembly line and disassembly task information, and establish a mixed integer linear programming mathematical model for the bilateral disassembly line balancing problem with the goal of minimizing the number of open workstations, idle time balance indicators, robot energy consumption and labor costs, including human-machine interaction constraint operations. The objective function is: F=min(f1,f2,f3,f4) Among them, minimize the number of open workstations: Minimize idle time balance indicator: Linearize it and get the minimum idle time balance index after linearization constraint: Where Q = |y iw -y jw | Minimize the energy consumption of the robotic workstation: Linearize it and get the energy consumption of the robot workstation minimized after linearization constraints: in, Minimize disassembly costs: Among them, S k is the workstation enable variable, indicating that if the kth workstation is enabled, then S k =1, otherwise S k =0; CT is the cycle time; I is the disassembly task set, {1,2,...,n}; M is the pairing workstation number set; L is the pairing station number, L={l|1,2}, l=1 means the left side, l=2 means the right side; W is the operator number, W={w|1,2}, w=1 means worker disassembly, w=2 means robot disassembly; i, j means disassembly task number; is a binary variable, indicating that if task i is assigned to the l-side of the k-th pairing station, then otherwise t i refers to the disassembly time of task i; OE refers to the robot's operating power consumption per unit time; refers to the operator assignment variable. If task i is assigned to w operators at the kth pairing station, then otherwise, Where y i1 Assign variables to the operator when the operator is a worker, y i2 The variable is assigned to the operator when the operator is a robot; SE refers to the robot's standby power consumption per unit time; SR kl Assign variables to the robot. If the l-side of the k-th paired station is a robot, then SR kl =1, otherwise, SR kl =0; C rt The unit time cost of robot disassembly; C mt is the unit time cost of workers when disassembling; BH is the set of tasks with hazardous attributes in the product to be disassembled; C h It refers to the additional cost required for dismantling hazardous tasks; And the constraints satisfied by the objective function are: Task allocation constraints: Operation allocation constraints: Time constraints: Precedence constraint: Workstation allocation constraints: The linearization constraints that need to be satisfied for linearization of the minimization of idle time balancing index are: Q ijw =max{and jw -and iw ,and iw -and jw } The linearization constraints that need to be satisfied to minimize the energy consumption of the robot workstation are: Binary variable range: In the above formulas, S k is the workstation enable variable, indicating that if the kth workstation is enabled, then S k =1, otherwise S k =0; CT is the cycle time; I is the disassembly task set, {1,2,...,n}; M is the pairing workstation number set; L is the pairing station number, L={l|1,2}, l=1 means the left side, l=2 means the right side; W is the operator number, W={w|1,2}, w=1 means worker disassembly, w=2 means robot disassembly; i, j means disassembly task number; is a binary variable, indicating that if task i is assigned to the l-side of the k-th pairing station, then otherwise t i refers to the disassembly time of task i; OE refers to the robot's operating power consumption per unit time; refers to the operator assignment variable. If task i is assigned to w operators at the kth pairing station, then otherwise, Where y i1 Assign variables to the operator when the operator is a worker, y i2 The variable is assigned to the operator when the operator is a robot; SE refers to the robot's standby power consumption per unit time; SR kl Assign variables to the robot. If the l-side of the k-th paired station is a robot, then SR kl =1, otherwise, SR kl =0; C rt The unit time cost of robot disassembly; C mt The unit time cost of workers when disassembling; C h Refers to the additional cost required to disassemble hazardous tasks; BI refers to the set of tasks that require interactive disassembly by workstations on both sides of the paired station in the product to be disassembled; BC refers to the set of tasks with complex attributes in the product to be disassembled; BH refers to the set of tasks with hazardous attributes in the product to be disassembled; i', j' are the numbers of the tasks that require system disassembly; T l Assign a set of tasks to each side of the paired station, if task i∈T l , then i can only be assigned to the l side of the paired station; is a positive number; ST i refers to the start time of task i, ST j Similarly; TP ij Refers to the priority relationship matrix of the disassembly task, TP ij =[a ij ] n×n ; if a ij =1, task i is the immediate predecessor of task j; is a disassembly sequence variable. If tasks i and j are assigned to w operators located on the l side of the k-th pairing station and task i is assigned before task j, then otherwise, Q ijw and u ijw All are variables required for constrained linearization; Step S2: Initialize the population, then sort and classify the population through fast non-dominant sorting, and establish an external archive to record the initial solution population after classification; Step S3: The classified initial solution population is optimized using the discrete squirrel search algorithm, and the final solution that meets the iteration conditions is output as the disassembly plan for execution.

2. The method for setting a bilateral disassembly line with human-machine interaction constraints according to claim 1, characterized in that: The population initialization step in step S2 includes encoding and decoding and calculating the target function to generate an initial population.

3. The method for setting a bilateral disassembly line with human-machine interaction constraints according to claim 1, characterized in that: The fast non-dominated sorting in step S2 is to classify the population based on the Pareto elite strategy.

4. The method for setting a bilateral disassembly line with human-machine interaction constraints according to claim 1, characterized in that: The position update process in the discrete squirrel search algorithm in step S3 performs discretized feature modification.

5. The method for setting a bilateral disassembly line with human-machine interaction constraints according to claim 1, characterized in that: In step S3, the seasonal constant in the discrete squirrel search algorithm is discretized and feature modified.

6. The method for setting a bilateral disassembly line with human-machine interaction constraints according to claim 1, characterized in that: The method for optimizing by the discrete squirrel search algorithm in step S3 is to optimize the individuals with better solution feasibility by the discrete squirrel search algorithm, update the external archive, and again use the fast non-dominant sorting based on the Pareto elite strategy to obtain the non-inferior solution that meets the iteration conditions in the external archive as the final result output.

Citation Information

Patent Citations

  • Space robot trajectory planning method based on sequence convex optimization

    CN113341731A

  • Arrangement of parallel maintenance lines for railway wagons

    US20200346675A1

  • Method for planning disassembly and reproduction of collected used products

    WO2006025082A1