Man-machine cooperation disassembly task allocation method and system considering disassembly complexity and task relevance

Abstract

The invention discloses a man-machine cooperation disassembly task allocation method and system considering disassembly complexity and task relevance. The method comprises the following steps: S1, collecting time required by a worker and a robot for completing each disassembly task and component attributes and disassembly process information of the worker for completing each disassembly task; s2, according to the information collected in the S1, an evaluation system of the disassembly task is established, and the evaluation system comprises SI score, disassembly complexity, task relevance, quantized disassembly cost, and physiological fatigue and psychological fatigue of workers; s3, based on the evaluation system determined in the S2, establishing a man-machine cooperation task allocation model which considers the disassembly complexity and the task relevance at the same time; and S4, solving the task allocation model proposed in the S3 by using an improved non-dominated sorting genetic algorithm, and outputting an optimal task allocation scheme. The capacity difference of workers and robots is comprehensively considered, the disassembly complexity is reduced, the working conditions of the workers are improved, and the disassembly quality is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of human-machine collaborative disassembly, and in particular relates to a human-machine collaborative disassembly task allocation method and system that considers disassembly complexity and task relevance. Background Art

[0002] With the rapid advancement of manufacturing technologies and the gradual shortening of product lifespans, the number of retired products is increasing. Consequently, the demand for remanufacturing of retired products, driven by both value and environmental considerations, is also growing. Disassembly, as the most critical step in separating obsolete products into their individual components through non-destructive or semi-destructive tasks, has gradually attracted attention. Unlike traditional manual or robotic disassembly, collaborative disassembly combines the capabilities of workers and robots within a shared workspace, achieving efficient disassembly by combining the strengths of both.

[0003] Due to the different characteristics of workers and robots, the disassembly complexity of the same task varies. Robots excel at performing repetitive, fixed tasks, while workers excel at flexible tasks that robots struggle with. Therefore, different task allocation strategies reflect different disassembly complexities. By properly considering and leveraging the complexity of disassembly tasks performed by both workers and robots, disassembly processes can be more effectively planned, resource utilization can be improved, and overall disassembly efficiency can be enhanced. Furthermore, the degree of inter-task correlation, as a key indicator reflecting the nature of tasks, plays a significant role in disassembly task planning. Failure to consider inter-task correlations can unavoidably negatively impact the performance of other tasks after detecting them. Therefore, understanding the nature of inter-task relationships is crucial for developing effective workflows, optimizing resource allocation, mitigating risks, and improving productivity. Furthermore, from the perspective of worker well-being, companies need to prioritize employee well-being. In addition to considering physical fatigue that affects workers' work status, psychological fatigue, as a factor affecting workers' psychological workload and attention, is even more important in the human-robot collaboration scenario where humans and robots interact closely. Therefore, task allocation should also take into account both physical and psychological fatigue of workers to accurately assess their actual work status, thereby reducing safety risks and the possibility of inefficiency in the workplace.

[0004] Existing task allocation technologies mainly focus on optimizing disassembly time and improving efficiency to meet the economic goals of enterprises. However, they still have the following shortcomings:

[0005] First, when measuring the complexity of disassembly tasks, existing techniques typically only quantitatively assess the performance of manual or robotic disassembly of a single object, or provide qualitative evaluations of both. However, there is a lack of simultaneous quantification of the complexity of both manual and robotic disassembly, taking into account the specific attributes of disassembly within a collaborative human-robot environment. This poses potential risks and efficiency challenges to task allocation and execution within collaborative human-robot collaboration. In this context, reducing disassembly complexity and improving disassembly quality by factoring in disassembly attributes has become a major challenge in the current disassembly process.

[0006] Second, existing techniques often focus solely on the characteristics and attributes of individual disassembly tasks, without considering the inter-task dependencies. This lack of consideration can have negative consequences, particularly when detecting anomalous tasks and their negative impact on the performance of other tasks cannot be avoided, thus reducing the flexibility and reliability of the disassembly process.

[0007] Third, most research on the human risk factors of collaborative disassembly has focused on the motion planning between workers and robots, such as gesture recognition, motion prediction, and collision avoidance. However, these studies often overlook the potential physical and mental fatigue that disassembly tasks can cause operators. This can lead to an inability to accurately assess workers' true working conditions during the disassembly process, increasing workplace safety risks and the potential for inefficiency. Summary of the Invention

[0008] In order to overcome the defects in the above-mentioned prior art, the present invention proposes a method and system for human-machine collaborative disassembly task allocation that takes into account disassembly complexity and task relevance. On the basis of ensuring disassembly efficiency, the present invention fully utilizes the characteristics of workers and robots from the perspectives of disassembly complexity, task relevance and human fatigue to optimize the task allocation of human-machine collaboration, in order to more comprehensively consider the ability differences between workers and robots, reduce disassembly complexity, improve workers' working conditions, and improve disassembly quality.

[0009] The present invention adopts the following technical solutions:

[0010] A method for allocating disassembly tasks in a human-machine collaborative manner, taking into account disassembly complexity and task relevance, comprises the following steps:

[0011] Step S1: Collect the time required for workers and robots to complete each disassembly task, as well as relevant information such as the component attributes and disassembly process of each disassembly task.

[0012] Step S2: Based on the information collected in step S1, an evaluation system is established for each disassembly task, including the Strain Index (SI) score, disassembly complexity, and inter-task dependencies of the task, while also quantifying the disassembly cost, physical fatigue, and psychological fatigue of workers.

[0013] Step S3: Based on the evaluation results determined in step S2, a human-machine collaborative task allocation model is established that takes both disassembly complexity and task relevance into consideration.

[0014] Step S4: Solve the task allocation model proposed in step S3 using the improved non-dominated sorting genetic algorithm and output the optimal task allocation solution.

[0015] Preferably, the evaluation of each disassembly task in step S2 is further described as follows:

[0016] For each disassembly task, an SI score was introduced to identify tasks that cause muscle damage to workers. The SI score was determined based on the subjective ratings of six task variables, including (1) exertion intensity (IE), (2) exertion duration per cycle (DE), (3) force per minute (EM), (4) hand / wrist posture (HWP), (5) work speed (SW), and (6) daily duration (DD).

[0017] Invite several technicians in the disassembly field to rate the SI variables of each disassembly task. According to the variable correspondence table (such as Example Table 1), the multipliers of each rating of the task variables can be found, and finally the SI score of each task is calculated according to the following formula.

[0018] SI=IE×DE×EM×HWP×SW×DD

[0019] Initial task assignment restrictions are derived based on the SI score. For example, all tasks with an SI score above 7 are considered harmful and are often associated with an increased risk of distal upper limb disorders. Therefore, tasks with an SI score greater than 7 will not be assigned to workers to ensure their health.

[0020] Preferably, the disassembly complexity attributes are divided into two categories: component attributes and connection attributes. Based on the U-rating data of manual and robotic disassembly, the average complexity coefficient of each component attribute and connection attribute feature is calculated (i.e., normalized). In the preferred embodiment, after optimization, the results are shown in Tables 2 and 3 of the embodiment.

[0021] After obtaining the average complexity coefficient of the corresponding attribute, the average component complexity coefficient and the average connection complexity coefficient of manual disassembly are calculated according to the following formula, where J1 and J2 are the number of component attributes and the number of connection attributes of manual disassembly (as shown in Table 2), respectively.

[0022]

[0023] The average component complexity coefficient and the average connection complexity coefficient of robot disassembly are calculated according to the following formula, where J3 and J4 are the number of component attributes and the number of connection attributes of robot disassembly (as shown in Table 3), respectively.

[0024]

[0025] The worker and robot disassembly complexity of each disassembly task are calculated using the following two formulas. is the worker's disassembly complexity for task i; is the disassembly complexity of the robot for task i; the superscripts H and R are used to distinguish the relevant parameters of the worker and the robot.

[0026]

[0027] Preferably, the task relevance assessment is quantified by the inter-task dependency, which is primarily represented by two indicators: sensitivity and variability. A fuzzy analytic hierarchy process (AHP) quantitative scale is used to quantify the two indicators. In a preferred embodiment, after optimization, the results are shown in Table 4.

[0028] Preferably, in step S2, the inter-task dependency is used to reflect the strength of task relevance, which includes sensitivity and variability;

[0029] Invite several technicians in the disassembly field to rate the sensitivity and variability of each task to determine the inter-task dependency of each disassembly task. The specific calculation formula is as follows. ij and V ij Represents the variability and sensitivity between task i and task j respectively; a ij It represents the dependency between task i and task j, and is used to reflect the strength of the correlation between tasks.

[0030]

[0031] The physical fatigue of workers when performing each task is mainly caused by long-term high-intensity physical labor during the disassembly process. In order to quantify physical fatigue, this paper adopts the LFFRM fatigue model. This model combines the learning-forgetting curve with the fatigue recovery model to describe the phenomenon that physical labor during the production process and production interruptions causes human fatigue, and rest relieves fatigue. The mathematical model of this model is applied to a periodic disassembly process. The specific calculation formula is as follows, where F i (t) represents the fatigue accumulation of t unit time when executing the i-th process, R(τ i ) indicates that after executing the i-th process, we rest for τ i The remaining fatigue amount after time, λ represents the physiological fatigue coefficient, and μ represents the worker's fatigue recovery coefficient.

[0032]

[0033] The mental fatigue of workers when performing each task is mainly caused by the mental load of concentrating attention, thinking and judging during the disassembly process. This paper uses the total information entropy of disassembly tasks within a certain period of time as an indicator of mental fatigue. The calculation method is shown in the following formula, where m i is the mental fatigue level corresponding to the i-th disassembly task performed by the worker (where m0 = 0), n i Indicates the number of transition states of the corresponding task during the operation process.

[0034]

[0035] In actual disassembly, since it takes a certain amount of time to digest mental fatigue, the present invention defines the mental fatigue degree of the current task as the weighted sum of the residual fatigue degree of the previous stage and the fatigue degree generated in the current stage. At the same time, the mental fatigue degree is controlled between the upper and lower limits of the normal threshold of human mental fatigue degree, as shown in the following formula. Where ρ represents the residual mental fatigue coefficient, I represents the number of tasks performed by the worker, and m i The meaning is the mental fatigue degree corresponding to the i-th disassembly task performed by the worker, m l and m u They respectively represent the lower and upper limits of the normal threshold of human psychological fatigue.

[0036]

[0037] For the evaluation of disassembly cost, the present invention divides the disassembly cost into labor cost, damage during robot disassembly, and loss of disassembly tools, and calculates it by the following formula. H 、P R and P represent the labor cost per second, robot maintenance cost per second, and power and disassembly tool loss costs, respectively. H 、T R and CT represent the time it takes for a worker to perform a task, the time it takes for a robot to perform a task, and the disassembly cycle time of a single product, respectively.

[0038] B=P H T H +P R T R +P·CT

[0039] Preferably, step S3 establishes a task allocation model that takes both disassembly complexity and task relevance into consideration, and is described in detail as follows:

[0040] Step S3-1: Assumptions of the model: (1) During the continuous disassembly of the same type of products, the task allocation plan for each product is the same; (2) Each task can be completed by the corresponding operator in a certain time; (3) When starting to disassemble the first product, the worker is not tired; (4) The worker's working time will not change with changes in factors such as worker fatigue; (5) The capabilities of resources of the same type are the same, and the heterogeneity of resources of the same type is not considered.

[0041] Step S3-2: Parameters and decision variables in the model: The decision variable is x ikt and y ilk , the former indicates that task i is executed by operator k at time t, which is 1, otherwise it is 0; the latter indicates that for operator k, if task l starts to be executed before task i, which is 1, otherwise it is 0. I represents the set of all disassembly tasks, I H and I R denotes the set of tasks performed by workers and robots respectively; K H and K R denote the set of all workers and the set of all robots respectively; and Represents the complexity of the disassembly task of the worker and the robot for task i, ID i represents the inter-task dependency of task i, T represents the set of discrete time periods within the task time range, t ik represents the time it takes for operator k to complete task i, P i represents the set of all preceding tasks corresponding to task i, μ represents the worker's physiological fatigue recovery coefficient; Z ijk If task i is assigned to the jth task executed by worker k, it is 1, otherwise it is 0; i and λ′ j The physiological fatigue coefficient of task i and the physiological fatigue coefficient of the worker assigned to perform task j, F max The safety threshold of physiological fatigue allowed for workers, m u and m l They represent the upper and lower limits of the safety threshold of the workers' mental fatigue, ρ represents the residual mental fatigue coefficient, and n i Indicates the number of transfer states of the corresponding task during the operation, m ik and m′ jk The mental fatigue of task i assigned to worker k and the mental fatigue of task j assigned to worker k, CT represents the disassembly time of unit product, TDC represents the overall disassembly complexity, and They represent the start and end time of worker k performing task j, and They represent the physiological fatigue of worker k at the start and end of task j, J k represents the number of tasks assigned to worker k, j represents the jth task performed by the worker, T H and T R They represent the worker working time and robot working time in a cycle, P H 、P R and P represent the labor cost per second, robot maintenance cost per second, and loss cost of electricity and disassembly tools, respectively. max It represents the maximum disassembly cost of a single product that the enterprise expects.

[0042] Step S3-3: Determine the model's objective functions. From an efficiency perspective, the goal is to minimize the disassembly cycle time (CT) for each product unit. Taking into account both task attributes and operator characteristics, the first objective is to minimize the total disassembly complexity (TDC) and the second is to minimize the inter-task dependencies (RD) of the tasks performed by the robot. Therefore, the three objective functions of the model are expressed as follows.

[0043] minF={CT,TDC,RD}

[0044] Step S3-4: Determine the constraints of the model based on the parameters obtained and established above. This step is further expanded as follows:

[0045] Each disassembly task can only be assigned to one worker or robot, expressed as:

[0046]

[0047] Each operator (including robots and workers) performs at most one task at any time, which can be expressed as:

[0048]

[0049] The actual disassembly time for each product is expressed as:

[0050]

[0051] The overall disassembly complexity of disassembling a product is expressed as:

[0052]

[0053] The inter-task dependency of the robot's execution tasks is expressed as:

[0054]

[0055] The execution of tasks must comply with the task priority relationship of the disassembly process, which is expressed as:

[0056]

[0057] Match the sequence number of each task with the order in which the task is assigned to the worker to perform, expressed as:

[0058]

[0059] Only when operator k completes its current job task can operator k continue to perform the next assigned job task, which is expressed as:

[0060]

[0061] Tasks that can only be completed by the corresponding operator can only be assigned to the corresponding operator, which is expressed as:

[0062]

[0063] The start and end times of worker k performing task j are expressed as:

[0064]

[0065] The physiological fatigue of worker k at the start and end of task j, and its control within the safety threshold, are expressed as:

[0066]

[0067] The calculation method of the mental fatigue of the disassembly task and its control within a reasonable threshold are expressed as:

[0068]

[0069] The physiological fatigue coefficient of task i and the psychological fatigue degree of task i are respectively mapped to the physiological fatigue coefficient and psychological fatigue degree of the jth worker performing the task, which can be expressed as:

[0070]

[0071] The time it takes for workers and robots to perform tasks, as well as the cost of disassembly, are within expectations, as shown below:

[0072]

[0073] P H T H +P R T R +P·CT≤B max

[0074] Preferably, step S4 uses an improved non-dominated sorting genetic algorithm to solve the proposed task allocation model and output the optimal task allocation solution. The further steps are as follows:

[0075] Step S4-1: Use double-layer real number encoding. The length of the chromosome is the number of tasks. The upper half of the chromosome represents the order in which the tasks are executed, and the lower half represents the executor of the corresponding task. First, randomly select a task from the set of unassigned tasks for which all previous tasks have been completed, and place the corresponding task number in the first position of the gene. Next, randomly select a task for which all previous tasks have been completed, and so on until the codes of all tasks are placed in the gene. For the operator assignment part, the value corresponding to each gene position is randomly generated from [1, h + r] (h represents the number of workers, r represents the number of robots).

[0076] Step S4-2: Population Initialization. According to the encoding rules, n individuals are randomly generated to form the initial population. Each individual represents a potential task allocation solution, including a task sequence and an operator allocation sequence. Decoding the encoding results of each individual yields a task allocation solution containing information such as the task execution order and operation time. The objective function values of each individual are then calculated using the three objective function calculation formulas, facilitating subsequent population optimization operations.

[0077] Step S4-3: Crossover and mutation. The two-point crossover method is used to cross the chromosomes corresponding to the tasks and operators respectively. The method is as follows: randomly select two crossover points and divide the parent chromosome into three parts. Then, delete the task number that is the same as the head and tail of the parent C1 from the parent C2, so that the remaining part of the parent C2 becomes the middle part of the child C1, and at the same time, the head and tail parts of the parent C1 are combined into the head and tail of the child C1. For the parent C1, a similar operation is performed to form the child C2. In this way, the child C1 and C2 still meet the task priority relationship after the crossover. The mutation process uses genetic mutation and structural mutation. Genetic mutation randomly replaces the executed operator with another one, while structural mutation directly swaps the positions of the two genes.

[0078] Step S4-4: To avoid falling into a local optimum, the present invention applies a simulated annealing algorithm to optimize individuals in the population. During the simulated annealing process, the current temperature is first initialized as the initial temperature, and then neighbor individuals of the current individual are continuously generated. Next, based on the task allocation scheme represented by the task sequence and operator allocation sequence in each individual, the three objective functions of the model are used as fitness values. The fitness values of the neighbor individuals are then calculated based on the collected information such as disassembly time and the calculation formula of the three objective functions. If the fitness value of the neighbor individual is better than the current best individual, the neighbor individual is accepted as the new current individual; if the fitness value is poor, the acceptance probability is calculated based on the current temperature and the fitness difference, and the poorer solution is accepted with a certain probability to escape the local optimum. This process is repeated until the current temperature drops to the minimum temperature or below.

[0079] Step S4-5: Evolution operation. The task sequence and operator sequence are recombined through crossover and mutation operations to generate n new individuals. After merging the populations, fast non-dominated sort is performed and the crowding distance of each individual is calculated. The elite individuals of the objective function are screened based on the results of fast non-dominated sort and the size of the crowding distance. The simulated annealing algorithm is then applied to optimize the new population.

[0080] Among them, the fast non-dominated sort is as follows:

[0081] For dominance relationship: If there are two individuals x1 and x2, and any objective function of x1 is better than the objective function of x2, then individual x1 is said to dominate individual x2, and the number of individuals i dominated by other individuals in the population is recorded as S i ;

[0082] ①Compare each objective function for all individuals in the population and find all S i = 1, assign these individuals a non-dominated rank of 1 and store them in the non-dominated set of rank = 1;

[0083] ② For each individual in the rank=1 set, the S of each individual in the set of individuals it dominates is i Subtract 1, if the individual's S i -1=0, then the individual is stored in the non-dominated set of rank=2;

[0084] ③ Repeat step ② in this way until all individuals are assigned non-dominated ranks, and the fast non-dominated sorting is completed;

[0085] The crowding distance is calculated as follows:

[0086] The crowding distance of an individual is used to reflect the superiority or inferiority of individuals in the same non-dominated level. The crowding distance of individual i is set to di , represents the value of individual i on the mth objective function (m = 1, 2, 3), i-1 and i+1 represent the left and right adjacent individuals of individual i, respectively. The calculation formula of the crowding distance is as follows:

[0087]

[0088] Step S4-6: Adjusting fatigue and cost constraints. Fatigue and cost thresholds are checked for individuals in the population, and task allocation is adjusted to ensure that workers' physical and mental fatigue and disassembly costs are within reasonable limits, while also meeting constraints such as task priority, the fact that disassembly tasks can only be performed by one operator, and that operators can only perform one task at a time.

[0089] Step S4-7: Iterative Update. After performing evolutionary operations and constraint adjustments on each generation of the merged population, n elite individuals are selected as the next generation population. The process then checks whether the maximum number of iterations, the termination condition, is met. If not, the iteration continues; if so, the process proceeds to step S4-8.

[0090] Step S4-8: Output Results. The first-level individuals from the final generation of the population, resulting from the fast non-dominated sorting, are used as the final Pareto solution set. Based on the task allocation scheme corresponding to each individual in the Pareto solution set, the disassembly time, total disassembly complexity, and inter-task dependencies of each individual are calculated. Finally, the task allocation scheme for each individual in the Pareto solution set and the corresponding three objective function values are output.

[0091] The present invention also discloses a human-machine collaborative disassembly task allocation system that considers disassembly complexity and task relevance, and implements the above method, including the following modules:

[0092] Information collection module: collects the time required for workers and robots to complete each disassembly task, as well as the component attributes and disassembly process information required for workers to complete each disassembly task;

[0093] Disassembly task evaluation system establishment module: Based on the collected information, an evaluation system for disassembly tasks is established. The evaluation system includes SI score, disassembly complexity and task relevance, as well as quantified disassembly cost, and workers' physical and mental fatigue.

[0094] Human-machine collaborative task allocation model establishment module: Based on a determined evaluation system, a human-machine collaborative task allocation model is established that takes into account both disassembly complexity and task relevance;

[0095] Optimal task allocation solution output module: uses the improved non-dominated sorting genetic algorithm to solve the task allocation model and output the optimal task allocation solution.

[0096] The present invention aims to improve the quality of human-machine matching in task allocation by analyzing the complexity of disassembly and the relevance of tasks, and to reduce the risk of worker fatigue and control disassembly costs, thereby protecting the workers' occupational health. In the present invention, paying attention to the complexity of disassembly and the relevance of tasks helps to allow workers and robots to perform disassembly tasks that they are good at, so as to achieve the vision of reducing errors in the disassembly process and improving the quality of disassembly, while paying attention to worker fatigue helps to improve workers' working conditions and ensure their well-being. To a certain extent, the present invention has promoted the innovation and development of human-machine collaborative disassembly technology, and has important practical application value for the development of the remanufacturing industry.

[0097] In summary, the significant technical effects of the present invention are specifically embodied in the following aspects:

[0098] 1. Considering the complexity of the disassembly task and reducing the overall disassembly complexity can avoid problems such as low disassembly quality due to high disassembly complexity for workers or robots, and increased errors in the disassembly process, which is conducive to improving product disassembly quality.

[0099] 2. Considering both physical and mental fatigue of workers during work in human-machine collaboration can more comprehensively assess the workload of workers during the disassembly process, thereby improving workers' well-being and working conditions.

[0100] 3. Incorporating the correlation between tasks into task allocation avoids the problem of unstable disassembly process caused by machines with low flexibility performing tasks with high task dependence, thereby improving the resilience of the disassembly process. BRIEF DESCRIPTION OF THE DRAWINGS

[0101] Figure 1 This is a diagram showing the priority relationship of the disassembly tasks of the Model 1sPBS provided in Example 1;

[0102] Figure 2 This is a diagram of a human-machine collaborative task allocation scheme obtained using a traditional method that only considers disassembly time in the human-machine collaborative mode of one person and one machine in Example 1;

[0103] Figure 3 This is a diagram of a human-machine collaborative task allocation scheme using the method of the present invention in the human-machine collaborative mode of one person and one machine in Example 1;

[0104] Figure 4 This is a diagram of a human-machine collaborative task allocation scheme using the method of the present invention in a human-machine collaborative mode of one person and two machines in Example 1;

[0105] Figure 5 This is a diagram of a human-machine collaborative task allocation scheme using the method of the present invention in a human-machine collaborative mode of two people per machine in Example 1;

[0106] Figure 6 This is a comparison chart of physiological fatigue corresponding to the human-machine collaborative task allocation scheme obtained using the method of the present invention and the traditional method that only considers disassembly time in the human-machine collaborative mode of one person and one machine in Example 1;

[0107] Figure 7 This is a comparison chart of psychological fatigue corresponding to the human-machine collaborative task allocation scheme obtained using the method of the present invention and the traditional method that only considers disassembly time in the human-machine collaborative mode of one person and one machine in Example 1;

[0108] Figure 8 This is a block diagram of a human-machine collaborative disassembly task allocation system that takes into account disassembly complexity and task relevance in Example 2. DETAILED DESCRIPTION

[0109] The preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0110] Example 1

[0111] This example uses Model 1s PBS as the research object. The power system of Model 1s PBS includes 16 battery modules, a battery management system (BMS) and necessary power electronic equipment. Figure 1 Model 1sPBS has a total of 22 disassembly tasks. The solid lines in the sequence diagram indicate mandatory precedence relationships, where the tasks at the end of the solid lines must be executed later than the tasks at the beginning of the lines. The dotted lines indicate non-mandatory precedence relationships. The task allocation scheme for this example is solved using the method of the present invention. The specific steps are as follows:

[0112] Step S1: Collect the time required for workers and robots to complete each task, as well as the component attributes and disassembly process (i.e., attached data) of each disassembly task completed by workers. Figure 1 Priority relationship of tasks in the

[0113] Step S2: Based on the information collected in step S1, an evaluation system is established for each disassembly task, including the SI score, disassembly complexity, and inter-task dependency of the task, while also quantifying the disassembly cost, physical fatigue, and mental fatigue of the workers.

[0114] Step S3: Based on the evaluation parameters determined in step S2, a human-machine collaborative task allocation model is established that takes both disassembly complexity and task relevance into consideration.

[0115] Step S4: Solve the task allocation model proposed in step S3 using the improved non-dominated sorting genetic algorithm and output the optimal task allocation solution.

[0116] In this embodiment, step S1 is used to obtain information such as the component attributes and disassembly process of the disassembly task. Disassembly work is performed in a human-robot collaborative work cell. There are three common modes of disassembly cells: one worker and one robot, one worker and two robots, and two workers and one robot.

[0117] In step S2, an evaluation system is established for each disassembly task, evaluating the SI scores of each task as shown in Table 1. The complexity of manual and robotic disassembly is then evaluated using the relevant attributes in Tables 2 and 3. Inter-task dependencies are then calculated using the variability and sensitivity indicators and scores in Table 4. Finally, the inter-task dependencies and the resulting physical and mental fatigue of workers are calculated for each disassembly task.

[0118] Table 1 SI indicator ratings and corresponding multiplier values

[0119]

[0120] Table 2 Disassembly attributes of manual disassembly

[0121]

[0122]

[0123] Table 3 Disassembly attributes of robot disassembly

[0124]

[0125]

[0126] Table 4. Metrics of inter-task dependencies

[0127]

[0128] After the above evaluation, we can obtain relevant information such as the disassembly complexity and task dependency of workers and robots for the 22 disassembly tasks, as shown in Table 5.

[0129] Table 5 Relevant information of disassembly task

[0130]

[0131]

[0132] In step S3, based on the above-mentioned known parameters, a human-machine collaborative task allocation model considering disassembly complexity and task relevance is established as follows:

[0133] The assumptions of the model are as follows: (1) During the continuous disassembly of the same type of products, the task allocation plan for each product is the same; (2) Each task can be completed by the corresponding operator in a certain time; (3) When starting to disassemble the first product, the worker is not tired; (4) The worker's working time will not change with changes in factors such as worker fatigue; (5) The capabilities of resources of the same type are the same, and the heterogeneity of resources of the same type is not considered.

[0134] Description of parameters and decision variables in the model: The decision variable is x ikt and y ilk , the former indicates that task i is executed by operator k at time t, which is 1, otherwise it is 0; the latter indicates that for operator k, if task l starts to be executed before task i, which is 1, otherwise it is 0. I represents the set of all disassembly tasks, I H and I R denote the set of tasks performed by workers and robots, K H and K R Represents the set of all workers and the set of all robots, and Represents the complexity of the disassembly task of the worker and the robot for task i, ID i represents the inter-task dependency of task i, T represents the set of discrete time periods within the task time range, t ik represents the time it takes for operator k to complete task i, P i represents the set of all preceding tasks corresponding to task i, μ represents the worker’s physiological fatigue recovery coefficient; Z ijk If task i is assigned to the jth task executed by worker k, it is 1, otherwise it is 0; i and λ′ j The physiological fatigue coefficient of task i and the physiological fatigue coefficient of the worker assigned to perform task j, F max The safety threshold of physiological fatigue allowed for workers, m u and m l They represent the upper and lower limits of the safety threshold of mental fatigue allowed for workers, ρ represents the residual mental fatigue coefficient, and n i Indicates the number of transfer states of the corresponding task during the operation, m ik and m′ jk The mental fatigue of task i assigned to worker k and the mental fatigue of task j assigned to worker k, CT represents the disassembly time of unit product, TDC represents the overall disassembly complexity, and They represent the start and end time of worker k performing task j, and They represent the physiological fatigue of worker k at the start and end of task j, J k represents the number of tasks assigned to worker k, j represents the jth task performed by the worker, T H and T R They represent the worker working time and robot working time in a cycle, P H 、P R and P represent the labor cost per second, robot maintenance cost per second, and loss cost of electricity and disassembly tools, respectively. max It represents the maximum disassembly cost of a single product that the enterprise expects.

[0135] The model has three objective functions. First, from an efficiency perspective, the goal is to minimize the disassembly cycle time (CT) for each product unit. Second, from the perspective of fully considering task attributes and operator characteristics, the first objective is to minimize the total disassembly complexity (TDC) and the second objective is to minimize the inter-task dependencies (RD) of the robot's tasks. These are expressed as follows.

[0136] minF={CT,TDC,RD}

[0137] The constraints of the model are as follows:

[0138] ① Each disassembly task can only be assigned to one worker or robot, expressed as:

[0139]

[0140] ② Each operator performs at most one task at any time, which can be expressed as:

[0141]

[0142] ③The actual disassembly time of each product is expressed as:

[0143]

[0144] ④ The overall disassembly complexity of a product is expressed as:

[0145]

[0146] ⑤The inter-task dependency of the robot is expressed as:

[0147]

[0148] ⑥The execution of the task must meet the requirements of the disassembly process before and after, which can be expressed as:

[0149]

[0150] ⑦ Match the sequence number of each task with the order in which the task is assigned to the worker to perform, expressed as:

[0151]

[0152] ⑧Only when operator k completes his current task can operator k continue to perform the next assigned task, which is expressed as:

[0153]

[0154] ⑨ Tasks that can only be completed by the corresponding operator can only be assigned to the corresponding operator, expressed as:

[0155]

[0156] ⑩ The start and end times of worker k performing task j are expressed as:

[0157]

[0158] The physiological fatigue of worker k at the start and end of task j, and its control within the safety threshold, are expressed as:

[0159]

[0160] The calculation method of the mental fatigue of the disassembly task and its control within a reasonable threshold are expressed as:

[0161]

[0162] The physiological fatigue coefficient of task i and the psychological fatigue degree of task i are respectively mapped to the physiological fatigue coefficient and psychological fatigue degree of the worker when performing the task, and expressed as:

[0163]

[0164] The time it takes for workers and robots to perform tasks, as well as the cost of disassembly, are within the company's expectations, as shown below:

[0165]

[0166]

[0167] P H T H +P R T R +P·CT≤B max

[0168] In step S4, based on the above optimization model, an improved non-dominated sorting genetic algorithm is used to solve the proposed task allocation model to obtain an optimal task allocation solution that meets the method proposed for the problem.

[0169] The above model is solved using an improved non-dominated sorting genetic algorithm. After parameter experimental setting, the population size is determined to be 300 and the number of iterations is determined to be 400. Finally, the following comparison table of Pareto solutions under different modes is obtained by randomly selecting from the obtained Pareto solutions.

[0170] Table 6 Comparison of Pareto solutions under different human-machine collaboration modes

[0171]

[0172] Table 6 shows the Pareto solutions obtained by the method of the present invention and the traditional method that only considers the minimization of disassembly time under three collaborative modes. Figure 2 as well as Figure 3 The table shows two examples of task allocation schemes for one person and one machine. Figure 4 as well as Figure 5 The table shows example diagrams of task allocation schemes obtained by this method under the modes of one person and two machines and two people and one machine. Analyzing the two diagrams under the mode of one person and one machine from the perspective of physiological fatigue, without considering the physiological fatigue constraint of the human body, that is, under the traditional allocation scheme, workers are often assigned more tasks, and during the entire disassembly process, workers only have a short break. When the physiological fatigue constraint is 0.8, workers have three longer break periods during the disassembly process. Figure 6 The physiological fatigue comparison chart shows that the peak physiological fatigue during the disassembly process designed by this method in the one-person-one-machine mode is 0.69, and the average physiological fatigue during the disassembly process is approximately between 0.3 and 0.4. Compared with traditional methods that do not consider physiological fatigue constraints, the method of this invention significantly reduces physiological fatigue of workers, thereby protecting their occupational health.

[0173] From the perspective of physiological fatigue, the method of the present invention can also significantly reduce the psychological fatigue of workers during the disassembly cycle through reasonable task planning, which can ensure that the disassembly workers are in good working condition. Figure 7As shown, under the traditional allocation scheme for the one-person-one-machine model, the peak psychological workload of workers reached approximately 12, exceeding the upper threshold of 10. However, under the allocation scheme obtained by the present invention, the peak psychological workload of workers was only approximately 5, a 58% reduction. Therefore, the present invention improves workers' psychological fatigue during task allocation, ensuring that the tasks assigned to them are appropriate. A reasonable task combination will keep workers' psychological fatigue within a reasonable range, effectively preventing a decline in their work performance and ensuring their efficiency.

[0174] From the perspective of the three objective functions in Table 6, although the task allocation model of the proposed method slightly increases the time required for disassembly, it effectively reduces the overall disassembly complexity and significantly reduces the inter-task dependencies among the tasks performed by the robot. This reduction in complexity allows both workers and robots to focus more on their respective specialized tasks, thereby reducing errors during the disassembly process. It also avoids the instability that can arise from less flexible machines performing highly inter-task dependencies, thereby improving the resilience of the disassembly process.

[0175] Example 2

[0176] like Figure 8 As shown, this embodiment discloses a human-machine collaborative disassembly task allocation system that considers disassembly complexity and task relevance, and executes the above method, including the following modules:

[0177] Information collection module: collects the time required for workers and robots to complete each disassembly task, as well as the component attributes and disassembly process information required for workers to complete each disassembly task;

[0178] Disassembly task evaluation system establishment module: Based on the collected information, an evaluation system for disassembly tasks is established. The evaluation system includes SI score, disassembly complexity and task relevance, as well as quantified disassembly cost, and workers' physical and mental fatigue.

[0179] Human-machine collaborative task allocation model establishment module: Based on a determined evaluation system, a human-machine collaborative task allocation model is established that takes into account both disassembly complexity and task relevance;

[0180] Optimal task allocation solution output module: uses the improved non-dominated sorting genetic algorithm to solve the task allocation model and output the optimal task allocation solution.

[0181] For other contents of this embodiment, please refer to the above method embodiment.

[0182] It should be noted that the above are only embodiments of the present invention and the technical principles adopted therein. Those skilled in the art will understand that the present invention is not limited to the above specific embodiments. Without exceeding the scope of protection of the present invention, those skilled in the art may make various obvious modifications, adjustments or substitutions. Therefore, although the present invention has been described in more detail above through specific embodiments, the present invention is not limited thereto and may also include other equivalent embodiments without departing from the inventive concept. The scope of protection of the present invention shall be subject to the appended claims.

Claims

1. A human-machine collaborative disassembly task allocation method considering disassembly complexity and task relevance is characterized by: The steps include: Step S1: Collect the time required for workers and robots to complete each disassembly task, as well as the component attributes and disassembly process information of each disassembly task; Step S2: Based on the information collected in step S1, an evaluation system for the disassembly task is established. The evaluation system includes SI score, disassembly complexity and task relevance, as well as quantified disassembly cost, worker physical fatigue and psychological fatigue; Step S3: Based on the evaluation system determined in step S2, a human-machine collaborative task allocation model is established that considers both disassembly complexity and task relevance; Step S4: Solve the task allocation model proposed in step S3 using the improved non-dominated sorting genetic algorithm and output the optimal task allocation solution.

2. The method for allocating disassembly tasks by human-machine collaboration taking into account disassembly complexity and task relevance as claimed in claim 1, characterized in that: In step S2, the SI score includes: (1) exertion intensity IE, (2) exertion duration per cycle DE, (3) exertion per minute EM, (4) hand or wrist posture HWP, (5) work speed SW, and (6) daily duration DD; Rating the SI variables of the disassembly task, finding the multiplier under each rating of the task variable, and calculating the SI score of the disassembly task according to the following formula; SI=IE×DE×EM×HWP×SW×DD Initial task assignment restrictions are derived based on the SI scores, where all tasks with an SI score higher than 7 are considered hazardous and disassembly tasks will not be assigned to workers.

3. The method for allocating disassembly tasks by human-machine collaboration taking into account disassembly complexity and task relevance as claimed in claim 2, characterized in that: In step S2, the disassembly complexity is divided into component attributes and connection attributes, and the average complexity coefficient of component attributes and connection attributes is calculated based on the U-rating data of manual disassembly and robot disassembly; After obtaining the average complexity coefficient of the corresponding attribute, the average component complexity coefficient and the average connection complexity coefficient of manual disassembly are calculated according to the following formula, where J1 and J2 are the number of component attributes and the number of connection attributes of manual disassembly, respectively; in, The meanings of are the average component complexity coefficient and the average connection complexity coefficient of manual disassembly respectively; The average component complexity coefficient and the average connection complexity coefficient of robot disassembly are calculated according to the following formula, where J3 and J4 are the number of component attributes and the number of connection attributes disassembled by the robot, respectively; in, The meanings of are the average component complexity coefficient and the average connection complexity coefficient of robot disassembly; The disassembly complexity of workers and robots for disassembly tasks is calculated using the following two formulas respectively; in, is the worker's disassembly complexity for task i; is the disassembly complexity of the robot for task i; the superscripts H and R are used to distinguish the relevant parameters of the worker and the robot.

4. The method for allocating disassembly tasks by human-machine collaboration taking into account disassembly complexity and task relevance as claimed in claim 3, characterized in that: In step S2, the inter-task dependency is used to reflect the strength of task relevance, including sensitivity and variability; The sensitivity and variability of the tasks are rated to determine the dependency between disassembly tasks. The specific formula is as follows; Among them, U ij and V ij Represent the variability and sensitivity between task i and task j, a ij represents the dependency between task i and task j; To quantify the physiological fatigue of workers while performing tasks, the LFFRM fatigue model is used. This model is applied to a cyclical disassembly process. The specific formula is as follows: F i (t)=R(τ i-1 )+(1-R(τ i-1 ))(1-e -λt ); Among them, F i (t) represents the fatigue accumulation of t unit time when executing the i-th process, R(τ i ) indicates a rest period τ after executing the i-th process i The remaining fatigue amount after time, λ represents the physiological fatigue coefficient, and μ represents the fatigue recovery coefficient of the worker; Regarding the psychological fatigue of workers when performing tasks, the total information entropy of processing disassembly tasks within a set time is used as an indicator of psychological fatigue. The calculation formula is as follows: Among them, m i is the mental fatigue level corresponding to the i-th disassembly task performed by the worker, n i Indicates the number of transition states of the corresponding task during the operation process; The mental fatigue level of the current task is defined as the weighted sum of the fatigue level left over from the previous stage and the fatigue level generated in the current stage. The mental fatigue level is controlled between the upper and lower limits of the normal threshold of mental fatigue for the human body, as shown in the following formula: Among them, m i The meaning is the mental fatigue degree corresponding to the i-th disassembly task performed by the worker, ρ represents the residual mental fatigue coefficient, I represents the number of tasks performed by the worker, m l and m u They represent the lower and upper limits of the normal threshold of human psychological fatigue respectively; The disassembly cost is divided into labor cost, damage to the robot during disassembly, and loss of disassembly tools, and is calculated using the following formula: B=P H T H +P R T R +P·CT Among them, P H 、P R and P represent the labor cost per second, robot maintenance cost per second, and disassembly tool loss cost, respectively. H 、T R and CT represent the time it takes for a worker to perform a task, the time it takes for a robot to perform a task, and the disassembly cycle time of a single product, respectively.

5. The method for allocating disassembly tasks by human-machine collaboration taking into account disassembly complexity and task relevance as claimed in claim 4, characterized in that: Step S3 is as follows: Step S3-1: Assumptions of the model: (1) During the continuous disassembly of the same type of products, the task allocation scheme for each product is the same; (2) Each task is completed within a certain time; (3) The worker is not fatigued when starting to disassemble the first product; (4) The worker's working time does not change; (5) The capabilities of the same type of resources are the same; Step S3-2: Parameters and decision variables in the model: The decision variable is x ikt and y ilk , the former indicates that if task i is executed by operator k at time t, it is 1, otherwise it is 0; the latter indicates that for operator k, if task l is started before task i, it is 1, otherwise it is 0; I represents the set of all disassembly tasks; I H and I R denotes the set of tasks performed by workers and robots respectively; K H and K R denote the set of all workers and the set of all robots respectively; and represents the complexity of the disassembly task of the worker and the robot for task i; ID i represents the inter-task dependency of task i; T represents the set of discrete time periods within the task time range; t ik represents the time it takes for operator k to complete task i; P i represents the set of all preceding tasks corresponding to task i; μ represents the worker’s physiological fatigue recovery coefficient; Z ijk If task i is assigned to the jth task executed by worker k, it is 1, otherwise it is 0; i and λ′ j The physiological fatigue coefficient of task i and the physiological fatigue coefficient of the worker assigned to perform task j respectively; F max Indicates the safety threshold of physiological fatigue allowed for workers; m u and m l Respectively represent the upper and lower limits of the safe threshold of mental fatigue allowed for workers; ρ represents the residual mental fatigue coefficient; n i Indicates the number of transition states of the corresponding task during the operation; m ik and m′ jk The mental fatigue of task i assigned to worker k and the mental fatigue of task j assigned to worker k are respectively; CT represents the disassembly time per unit product; TDC represents the overall disassembly complexity; and They represent the start and end times of worker k performing the jth task respectively; and They represent the physiological fatigue of worker k at the start and end of task j; J k represents the number of tasks assigned to worker k; j represents the jth task performed by the worker; T H and T R They represent the worker working time and the robot working time in one cycle respectively; H 、P R and P represent the labor cost per second, robot maintenance cost per second, and loss cost of electricity and disassembly tools respectively; B max It represents the maximum disassembly cost of a single product expected by the enterprise; Step S3-3: Determine the objective function of the model as follows; minF={CT,TDC,RD} Among them, CT represents the disassembly cycle time of each product unit, TDC represents the total disassembly complexity, and RD represents the inter-task dependency of the robot's execution tasks; Step S3-4: Determine the constraints of the model.

6. The method for allocating disassembly tasks by human-machine collaboration taking into account disassembly complexity and task relevance as claimed in claim 5, characterized in that: Steps S3-4 are as follows: Each disassembly task can only be assigned to one worker or robot, expressed as: Each operator performs at most one task at any time, which can be expressed as: The actual disassembly time for each product is expressed as: The overall disassembly complexity of disassembling a product is expressed as: The inter-task dependency of the robot's execution tasks is expressed as: The execution of tasks complies with the task priority relationship of the disassembly process, which is expressed as: Match the sequence number of each task with the order in which the task is assigned to the worker to perform, expressed as: Only when operator k completes its current job task can operator k continue to perform the next assigned job task, which is expressed as: The tasks completed by the corresponding operators can only be assigned to the corresponding operators, which is expressed as: The start and end times of worker k performing task j are expressed as: The physiological fatigue of worker k at the start and end of task j, and the physiological fatigue at the end of task j controlled within the safety threshold, are expressed as: The calculation formula for the mental fatigue of the disassembly task and its control within a reasonable threshold are expressed as: The physiological fatigue coefficient of task i and the psychological fatigue degree of task i are respectively mapped to the physiological fatigue coefficient and psychological fatigue degree of the jth worker performing the task, which can be expressed as: The time it takes for workers and robots to perform tasks, as well as the cost of disassembly, are within expectations, as shown below: P H T H +P R T R +P·CT≤B max 。 7. The method for allocating disassembly tasks by human-machine collaboration taking into account disassembly complexity and task relevance as claimed in claim 6, characterized in that: Step S4 is specifically as follows: Step S4-1: Use double-layer real number encoding; the length of the chromosome is the number of tasks, the upper half of the chromosome represents the order in which the tasks are executed, and the lower half represents the executor of the corresponding task; first, randomly select a task from the set of unassigned tasks for which all previous tasks have been completed, and place the corresponding task number in the first position of the gene; second, randomly select a task for which all previous tasks have been completed, and so on until the codes of all tasks are placed in the gene; for the operator assignment part, the value corresponding to each gene position is randomly generated from [1, h+r]; h represents the number of workers, and r represents the number of robots; Step S4-2: Population initialization: n individuals are randomly generated according to the encoding rules to form an initial population. Each individual represents a potential task allocation scheme, including a task sequence and an operator allocation sequence. The encoding result of each individual is decoded to obtain a task allocation scheme containing task execution order and operation time information. The objective function value of each individual is calculated using the three objective function calculation formulas. Step S4-3: Crossover and mutation. Using a two-point crossover method, the chromosomes corresponding to the tasks and operators are crossed separately, as follows: Two crossover points are randomly selected to divide the parent chromosome into three parts; the task number that is the same as the head and tail of the parent C1 is deleted from the parent C2, so that the remaining part of the parent C2 becomes the middle part of the offspring C1, and the head and tail of the parent C1 are combined to form the head and tail of the offspring C1; for the parent C1, the same operation as the parent C2 is performed to form the offspring C2; the mutation process uses genetic mutation and structural mutation. Genetic mutation randomly replaces the executed operator with another one, and structural mutation directly swaps the positions of two genes. Step S4-4: Apply the simulated annealing algorithm to the individuals in the optimization population; during the simulated annealing process, first initialize the current temperature as the initial temperature, and then continuously generate neighbor individuals of the current individual; then, based on the task allocation scheme represented by the task sequence and operator allocation sequence in each individual, use the three objective functions of the model as fitness values, and then calculate the fitness values of the neighbor individuals based on the collected disassembly time information and the calculation formulas of the three objective functions in the model. If the fitness value of the neighbor individual is better than the current best individual, then accept the neighbor individual as the new current individual; if the fitness value is lower than the set value, calculate the acceptance probability based on the current temperature and the difference between the objective function value of the neighbor individual and the objective function value of the current individual, and accept the solution with the set probability; repeat until the current temperature drops to or below the set minimum temperature; Step S4-5: Evolution operation: Recombining the task sequence and operator sequence through crossover and mutation operations to generate n new individuals. After merging the population, perform fast non-dominated sorting and calculate the crowding distance of each individual. Then, based on the results of fast non-dominated sorting and the size of the crowding distance, screen out the elite individuals of the objective function, and then use the simulated annealing algorithm to optimize the new population. Step S4-6: Fatigue and cost constraint adjustment: Fatigue and cost thresholds are checked for individuals in the population, and task allocation is adjusted to ensure that the worker's physical and mental fatigue and disassembly cost are within the set range, and that the constraints of task priority, the disassembly task can only be performed by one operator, and the operator can only perform at most one task at a time are met; Step S4-7: Iterative update; after performing evolutionary operations and constraint adjustments on the merged population of each generation, n elite individuals can be selected as the next generation population, and then the termination condition of the maximum number of iterations is checked to see whether it is met; If not, continue the iteration; if satisfied, execute step S4-8; Step S4-8: output the result; The first-level individuals of the last generation of population executing the fast non-dominated sorting result are taken as the final Pareto solution set. Based on the task allocation plan corresponding to each individual in the obtained Pareto solution set, the disassembly time, total disassembly complexity and inter-task dependency of each individual are obtained; finally, the task allocation plan and the corresponding three objective function values of each individual in the Pareto solution set are output.

8. The method for allocating disassembly tasks by human-machine collaboration taking into account disassembly complexity and task relevance as claimed in claim 7, characterized in that: In step S4-5, the fast non-dominated sort is specifically as follows: For dominance relationship: If there are two individuals x1 and x2, and any objective function of x1 is better than the objective function of x2, then individual x1 is said to dominate individual x2, and the number of individuals i dominated by other individuals in the population is recorded as S i ; ①Compare each objective function for all individuals in the population and find all S i = 1, assign these individuals a non-dominated rank of 1 and store them in the non-dominated set of rank = 1; ② For each individual in the rank=1 set, the S of each individual in the set of individuals it dominates is i Subtract 1, if the individual's S i -1=0, then the individual is stored in the non-dominated set of rank=2; ③ Repeat step ② until all individuals are assigned non-dominated ranks, and the fast non-dominated sorting is completed; The crowding distance is calculated as follows: Set the crowding distance of individual i to d i , represents the value of individual i on the mth objective function, m = 1, 2, 3, i-1 and i+1 represent the left and right adjacent individuals of individual i, respectively. The calculation formula of the crowding distance is as follows:

9. A human-machine collaborative disassembly task allocation system considering disassembly complexity and task relevance, executing the method according to any one of claims 1 to 8, characterized in that: Includes the following modules: Information collection module: collects the time required for workers and robots to complete each disassembly task, as well as the component attributes and disassembly process information required for workers to complete each disassembly task; Disassembly task evaluation system establishment module: Based on the collected information, an evaluation system for disassembly tasks is established. The evaluation system includes SI score, disassembly complexity and task relevance, as well as quantified disassembly cost, and workers' physical and mental fatigue. Human-machine collaborative task allocation model establishment module: Based on a determined evaluation system, a human-machine collaborative task allocation model is established that takes into account both disassembly complexity and task relevance; Optimal task allocation solution output module: uses the improved non-dominated sorting genetic algorithm to solve the task allocation model and output the optimal task allocation solution.

Images