Workshop scheduling method and apparatus considering skill levels and fatigue degrees of workers

Abstract

A workshop scheduling method and apparatus considering skill levels and fatigue degrees of workers. The method comprises: on the basis of production order data, worker data and workshop device data, constructing a production scheduling example by means of a workshop scheduling model; for each scheduling operation, determining state parameters of a workshop on the basis of the production scheduling example and a scheduling Gantt chart; determining a search strategy used by the scheduling operation; when the search strategy is random search, randomly selecting a scheduling rule from among a plurality of scheduling rules as the current scheduling rule; when the search strategy is network prediction, inputting the state parameters of the workshop into a scheduling network, so that the scheduling network determines a reward value of each scheduling rule among the plurality of scheduling rules corresponding to the state parameters, and uses the scheduling rule with the maximum reward value as the current scheduling rule; on the basis of the current scheduling rule, determining a scheduling group corresponding to the scheduling operation; and decoding the scheduling group to the scheduling Gantt chart by using a greedy decoding mode.

Description

Workshop scheduling method and device considering skill level and fatigue degree of workers TECHNICAL FIELD

[0001] The present application relates to the technical field of workshop scheduling, and particularly relates to a workshop scheduling method and device considering skill level and fatigue degree of workers. BACKGROUND

[0002] With the diversification and individualization of market demand, manufacturing enterprises are facing the challenge of more complex production processes. In the workshop, production resources such as equipment, manpower and raw materials are limited, which leads to tension in production planning and increase in downtime and waiting time in the production process. This situation may cause a decline in production efficiency and problems. Therefore, manufacturing enterprises need to arrange limited production resources through workshop scheduling to improve production efficiency. SUMMARY

[0003] Therefore, the present application provides a workshop scheduling method and device considering skill level and fatigue degree of workers, which arranges limited production resources through workshop scheduling to improve production efficiency.

[0004] Specifically, the present application is realized through the following technical solutions:

[0005] The first aspect of the present application provides a workshop scheduling method considering skill level and fatigue degree of workers, the workshop scheduling method comprising:

[0006] constructing a workshop scheduling model, and constructing a production scheduling example based on the workshop scheduling model according to production order data, worker data and workshop equipment data; wherein the workshop scheduling model is a workshop scheduling model considering skill level and fatigue degree of workers;

[0007] solving the production scheduling example by using a scheduling algorithm to obtain a scheduling solution of the production scheduling example;

[0008] wherein solving the production scheduling example by using a scheduling algorithm comprises:

[0009] for each scheduling operation, determining the state parameters of the workshop at the current time based on the production scheduling example and the scheduling Gantt chart at the current time;

[0010] determining the search strategy used for the current scheduling operation;

[0011] when the search strategy is random search, a scheduling rule is randomly selected from a plurality of preset scheduling rules as a current scheduling rule; when the search strategy is network prediction, a current state parameter of the workshop is input into a scheduling network, so as to determine a reward value corresponding to each scheduling rule in the plurality of scheduling rules by the scheduling network according to the state parameter, and a scheduling rule with the maximum reward value is taken as the current scheduling rule;

[0012] a scheduling group corresponding to the current scheduling rule is determined according to the current scheduling rule; wherein the scheduling group is a combination of workpieces, machines and workers;

[0013] the scheduling group is decoded into a scheduling Gantt chart by using a greedy decoding mode.

[0014] The second aspect of the application provides a workshop scheduling device considering skill levels and fatigue degrees of workers, the device comprising a construction module and a solving module; wherein,

[0015] the construction module is configured to construct a workshop scheduling model, and construct a production scheduling example based on the workshop scheduling model according to production order data, worker data and workshop equipment data; wherein the workshop scheduling model is a workshop scheduling model considering skill levels and fatigue degrees of workers;

[0016] the solving module is configured to solve the production scheduling example by using a scheduling algorithm, so as to obtain a scheduling solution of the production scheduling example;

[0017] wherein solving the production scheduling example by using the scheduling algorithm comprises:

[0018] for each scheduling operation, a state parameter of the workshop at a current time is determined based on the production scheduling example and a scheduling Gantt chart at the current time;

[0019] a search strategy used for the current scheduling operation is determined;

[0020] when the search strategy is random search, a scheduling rule is randomly selected from a plurality of preset scheduling rules as a current scheduling rule; when the search strategy is network prediction, a current state parameter of the workshop is input into a scheduling network, so as to determine a reward value corresponding to each scheduling rule in the plurality of scheduling rules by the scheduling network according to the state parameter, and a scheduling rule with the maximum reward value is taken as the current scheduling rule;

[0021] a scheduling group corresponding to the current scheduling rule is determined according to the current scheduling rule; wherein the scheduling group is a combination of workpieces, machines and workers;

[0022] the scheduling group is decoded into a scheduling Gantt chart by using a greedy decoding mode.

[0023] The third aspect of the present application provides a workshop scheduling device considering the skill level and fatigue degree of workers, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any of the methods provided in the first aspect of the present application when executing the program.

[0024] The fourth aspect of the present application provides a computer readable storage medium, which stores a computer program, wherein the program implements the steps of any of the methods provided in the first aspect of the present application when executed by a processor.

[0025] The method and device for workshop scheduling considering the skill level and fatigue degree of workers provided by the present application have the following advantages. In the first aspect, the workshop scheduling is achieved by constructing a workshop scheduling model and a production scheduling example and solving the production scheduling example by the workshop scheduling model, so that the production resources can be reasonably arranged, the production process can be optimized, and the production efficiency of the workshop is improved. In the second aspect, the skill level, fatigue degree and system carbon emission of workers are considered when the workshop scheduling model is constructed, so that the carbon emission generated in production can be reduced under the condition of maintaining the production fairness of workers, which can help manufacturing enterprises to establish a more scientific and sustainable workshop scheduling model, improve the production efficiency, and reduce the carbon emission, and is of great significance to the sustainable development of enterprises. In the third aspect, the production scheduling example is solved by using different search strategies, so that the generalization of the scheduling solution is improved, the production benefit of the workshop is improved, real-time scheduling solution is achieved, the solving speed of the scheduling solution is greatly improved, and the scheduling speed is improved. BRIEF DESCRIPTION OF DRAWINGS

[0026] FIG. 1 is a flowchart of the method for workshop scheduling considering the skill level and fatigue degree of workers provided by the present application;

[0027] FIG. 2 is a schematic diagram of a workshop scheduling model according to an example embodiment of the present application;

[0028] FIG. 3 is a flowchart of solving a production scheduling example based on a scheduling algorithm according to an example embodiment of the present application;

[0029] FIG. 4 is a schematic diagram of a scheduling Gantt chart according to an example embodiment of the present application;

[0030] FIG. 5 is a flowchart of the method for workshop scheduling considering the skill level and fatigue degree of workers according to an example embodiment of the present application;

[0031] FIG. 6 is a schematic diagram of the implementation principle of a greedy decoding according to an example embodiment of the present application;

[0032] FIG. 7 is a flowchart of the method for workshop scheduling considering the skill level and fatigue degree of workers provided by the present application;

[0033] Fig. 8 is a schematic diagram of an implementation of training a scheduling network according to an example embodiment of the present application;

[0034] Fig. 9 is a schematic diagram of a hardware structure of a workshop scheduling device according to an example embodiment of the present application;

[0035] Fig. 10 is a schematic diagram of a structure of a first example embodiment of a workshop scheduling device according to an example embodiment of the present application; DETAILED DESCRIPTION

[0036] The example embodiments will be described in detail herein with reference to the drawings. When the description below refers to accompanying drawings, unless otherwise specified, the same numbers in different drawings refer to the same or similar elements. The implementations described in the following example embodiments are not meant to represent all implementations consistent with the present application.

[0037] The terminology used in the present application is for the purpose of describing particular example embodiments only and is not intended to be limiting of the present application. As used in the present application, the singular forms "a," "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0038] It will be understood that, although the terms first, second, third, etc. can be used herein to describe various information, these terms are not intended to denote a particular order or hierarchy among the information. These terms are used merely as labels to identify particular information. For example, a first information can be termed a second information, and similarly, a second information can be termed a first information without departing from the scope of the present application. As used herein, the word "if' can be interpreted to mean "when" or "upon" or "in response to determining" taking into account the context in which the term is used.

[0039] The following specific examples are given to provide a detailed description of the technical solutions of the present application.

[0040] Fig. 1 is a flowchart of a first example embodiment of a workshop scheduling method according to the present application.

[0041] Referring to Fig. 1, the workshop scheduling method according to the present embodiment can include:

[0042] S101, constructing a workshop scheduling model, and based on production order data, worker data and workshop equipment data, constructing a production scheduling example based on the workshop scheduling model; wherein the workshop scheduling model is a workshop scheduling model considering the skill level and fatigue degree of workers.

[0043] Specifically, the workshop scheduling model is used to construct a production scheduling example according to workshop data (the workshop includes production order data, worker data and workshop equipment data), which considers the skill level and fatigue degree of the workers.

[0044] It should be noted that the workshop scheduling model is constructed according to actual needs, and FIG. 2 is a schematic diagram of a workshop scheduling model according to an example embodiment of the present application. Please refer to FIG. 2, in the example of the present embodiment, the workshop scheduling model constructed can be described as follows: in a production system, there are n workpieces to form a workpiece task set J=(J1, J2, …, Jn), a set W=(W1, W2, …, Ww) of w workers processes on a machine set M=(M1, M2, …, Mm) including m machines. Each workpiece Ji has a fixed order of ri process OPi=(Oii, Oi2, …, Oir), the process Oij indicates that the jth process of the workpiece i can be processed on a specific subset of the machine set M, each worker Wa has a specific skill set Worker skill refers to the worker's ability to operate a specific machine, and each worker has a one-to-one corresponding skill level set When the equipment type is an automated processing equipment, the quantification of the skill level is the initial operation time of the worker Wa operating the machine Mp; when the equipment is a man-machine collaborative processing equipment, the quantification of the skill level is the proportion of the first operation of the worker Wa operating the machine Mp exceeding the standard working hours; in each step of processing assignment, only when the workpiece Ji, the machine Mp and the worker Wa are ready at the same time, the processing task can be performed. In other words, the workshop scheduling model can be divided into three sub-problems: (1) workpiece selection: determine the workpiece processing order in each batch processing task; (2) machine allocation: determine the machine allocation of each workpiece for each process; (3) worker assignment: determine the operating worker of each process on the machine.

[0045] Further, the objective function of the workshop scheduling model includes a first objective function for minimizing the total production time, a second objective function for minimizing the system carbon emission and a third objective function for minimizing the standard deviation of worker fatigue; wherein,

[0046] The first objective function is:

[0047] Wherein, the TE is the total production time; the is the total processing time of the jth process of the workpiece i; the is the skill level of the worker operating the machine; the H is the incompressibility coefficient; the is the number of times of the worker Wa operating the machine Mp; the learning rate of worker Wa; the base processing time of the jth process of workpiece i on machine Mp; the MKm is a machine type;

[0048] The second objective function is: f2 = min C E C E = C AE +C CM +C MR +C TR +C FP

[0049] wherein the C E is system carbon emission; the C AE is total carbon emission of auxiliary processing instrument loss; the C CM is total carbon emission generated due to use of processing consumables; the C MR is total carbon emission of equipment operation of manufacturing system for completing manufacturing task; the C TR is total carbon emission of carrying in processing; the C FP is total carbon emission of workshop public equipment;

[0050] The third objective function is:

[0051] wherein the VF E is worker fatigue standard deviation; the is cumulative fatigue value of worker W a at t+1; the W is the number of workers in the production system; the is decision variable of worker working state; the R t is workload coefficient defined according to machine type operated by the worker; the is cumulative fatigue value of worker W a at t; the is fatigue accumulation rate determined by personal physical quality of the worker; the is fatigue recovery rate determined by personal physical quality of the worker;

[0052] The constraint condition of workshop scheduling model is:

[0053] wherein the m represents the number of machines; the p represents the serial number of machines; the represents whether the process O i of workpiece J ij is processed on machine M p ; the represents whether the process O ithe process O ij whether the worker W a processing; the w represents the number of workers; the a represents the sequence number of the worker; the whether the worker W a whether the worker W p the use method; the j, k represent the sequence number of the process; the the start processing time of the jth process of the workpiece i; the the end processing time of the jth process of the workpiece i.

[0054] In other words, the constraint conditions of the workshop scheduling model are: ensuring that there is only one machine processing for each process of each workpiece; ensuring that there is only one worker processing for each process of each workpiece; ensuring that each machine is used by at least one worker; ensuring that the worker only operates the machine that the worker can operate for processing; ensuring the sequence constraint between the processes of each workpiece, i.e., the start processing time of the subsequent process cannot be earlier than the end processing time of the previous process.

[0055] Further, the production order data includes the workpiece name, the process number, the process set, the process optional machine set, the process optional machine processing basic time, the workpiece material, the workpiece quality, the workpiece yield, and the material utilization rate. The worker data includes the worker name, the worker skill set, the worker first operation time, the learning rate, the fatigue accumulation rate, and the fatigue recovery rate. The workshop equipment data includes the machine name, the man-machine interaction type, the idle power, the load power, the auxiliary processing instrument set, the processing instrument life, the consumable set, the unit time consumable amount, and the machine position.

[0056] In the specific implementation, the calculated production order data, worker data, and workshop equipment data are input into the constructed workshop scheduling model in real time, and thus the production scheduling example can be obtained.

[0057] S102, solving the production scheduling example by using a scheduling algorithm to obtain a scheduling solution of the production scheduling example.

[0058] Specifically, the scheduling algorithm is used to solve the production scheduling example, and the scheduling solution obtained is used to control the scheduling group (combination of workpieces, machines, and workers) of the workshop. FIG. 3 is a flowchart of solving the production scheduling example by using a scheduling algorithm according to an example embodiment of the present application. Please refer to FIG. 3. The process of solving the production scheduling example by using a scheduling algorithm can include:

[0059] S301, determining the state parameters of the workshop at the current time based on the production scheduling example and the scheduling Gantt chart at the current time for each scheduling operation.

[0060] Specifically, the scheduling Gantt chart is used to visualize the progress and time arrangement of the workshop scheduling. FIG. 4 is a schematic diagram of a scheduling Gantt chart according to an example embodiment of the present application.

[0061] Further, the state parameters represent the state of the workshop (the state of the workpieces, machines and workers), which includes the average workpiece completion rate, the workpiece completion rate standard deviation, the expected completion time, the system expected total discharge, the discharge standard deviation, the average machine effective work discharge rate, the machine effective work discharge rate standard deviation, the worker average fatigue, and the worker fatigue standard deviation.

[0062] It should be noted that the state parameters can be normalized state parameters. It can be understood that each state parameter includes a corresponding influence factor. The influence factor of the expected completion time is the average of the sum of the maximum processing time of all processes and the sum of the minimum processing time, the influence factor of the system expected total discharge is the number of machines, the average machine processing discharge, the average machine idle discharge, the average of the sum of the maximum processing time of all processes and the sum of the minimum processing time, the influence factor of the discharge standard deviation is the total system discharge under the current state, the influence factor of the worker average fatigue is the average of the sum of the maximum processing time of all processes and the sum of the minimum processing time, and the influence factor of the worker fatigue standard deviation is the worker average fatigue.

[0063] The normalized state parameter can be represented as: pf= pf s / Пτ

[0064] Wherein, the pf is the normalized state parameter; the pf s is the state parameter before normalization; and the ∏ τ is the product of the influence factors corresponding to all state parameters.

[0065] In a specific implementation, for each scheduling operation, the state parameters are determined based on the production scheduling example and the scheduling Gantt chart.

[0066] S302, determine the search strategy used in the current scheduling operation.

[0067] Specifically, the search strategy includes random search and network prediction. The specific implementation and implementation process of random search and network prediction will be described below, and will not be described here.

[0068] In a specific implementation, the search strategy used in the current scheduling operation can be determined by an ε-greedy strategy. That is, a number δ is randomly generated, and the relationship between the random number δ and the current ε is determined. If δ is greater than ε, it is determined that a random search is used. If δ is less than ε, it is determined that network prediction is used. It should be noted that ε is set according to actual needs, and in this embodiment, the specific value of ε is not limited. For example, in an embodiment, the initial value of ε is 0.80, and ε is updated with the number of iterations. The value of ε decreases by 0.001 every iteration.

[0069] S303, when the search strategy is a random search, randomly selecting one of the plurality of preset scheduling rules as the current scheduling rule; when the search strategy is network prediction, inputting the current state parameter of the workshop into the scheduling network to determine the reward value corresponding to each of the plurality of scheduling rules by the scheduling network according to the state parameter, and selecting the scheduling rule with the maximum reward value as the current scheduling rule.

[0070] Specifically, the plurality of preset scheduling rules are set according to actual needs, and in this embodiment, this is not limited. For example, in an embodiment, the plurality of preset scheduling rules include: 1. {FIFO}, 2. {FOFI}, 3. {LPT}, 4. {SPT}, 5. {MCO}, 6. {LTH}, 7. {LCR, LCE, LTW}, 8. {LCR, LCE, SOT}, 9. {LCR, SMT, LTW}, 10. {LCR, SMT, SOT}, 11. {HCR, LCE, LTW}, 12. {HCR, LCE, SOT}, 14. {HCR, SMT, LTW}, 14. {HCR, SMT, SOT};

[0071] The FIFO indicates selecting a workpiece-machine-worker scheduling group that can be processed earliest;

[0072] The FOFI indicates selecting a workpiece-machine-worker scheduling group that can complete processing earliest;

[0073] The LPT indicates selecting a workpiece-machine-worker scheduling group with the longest processing time;

[0074] The SPT indicates selecting a workpiece-machine-worker scheduling group with the shortest processing time;

[0075] The MCO indicates selecting a workpiece-machine-worker scheduling group that increases the least emissions;

[0076] The LTH indicates selecting a workpiece-machine-worker scheduling group that makes the least fatigue of the worker accumulate the most fatigue;

[0077] The LCR indicates selecting a workpiece with the smallest completion rate.

[0078] The HCR represents selecting a workpiece with the largest completion rate;

[0079] The LCE represents selecting a machine with the smallest carbon emission;

[0080] The SMT represents selecting a machine with the shortest processing time;

[0081] The LTW represents selecting a worker with the smallest current fatigue;

[0082] The SOT represents selecting a worker with the shortest operation time.

[0083] The FIFO, the FOFI, the LPT, the SPT, the MCO and the LTH are composite heuristic rules, the LCR and the HCR are workpiece scheduling rules, the LCE and the SMT are machine scheduling rules, and the LTW and the SOT are worker scheduling rules.

[0084] Optionally, in a possible implementation, the scheduling network is configured to determine a current scheduling rule corresponding to a state parameter of the workshop at a current time, and the scheduling network includes a first model TDQN (Temporal Difference Deep Q-Network) and a second model ADQN (Advantage Deep Q-Network). The first model inputs the state parameter and outputs a current decision tendency under the state parameter. The second model inputs the state parameter and the current decision tendency and outputs the current scheduling rule.

[0085] In a specific implementation, FIG. 5 is a flowchart of a second embodiment of a workshop scheduling method considering worker skill levels and fatigue degrees according to an example embodiment of the present application. The method provided in this embodiment, when the scheduling network includes the first model and the second model, the scheduling network determines reward values of each of the scheduling rules corresponding to the state parameter, and the scheduling rule with the largest reward value is determined as the current scheduling rule, and the method includes the following steps.

[0086] S501, the first model determines reward values of a plurality of preset decision tendencies corresponding to the state parameter, and determines the decision tendency with the largest reward value as the current decision tendency.

[0087] Specifically, the plurality of preset decision tendencies represent the current optimization target of the workshop scheduling model, which is set according to actual needs, and is not limited in the embodiment. For example, in an embodiment, the decision tendency includes three values, which are 0, 1 and 2 respectively. Among them, the decision tendency of 0 represents that the workshop scheduling model is biased to optimize the total production time, the decision tendency of 1 represents that the workshop scheduling model is biased to optimize the system carbon emission, and the decision tendency of 2 represents that the workshop scheduling model is biased to optimize the worker fatigue standard deviation.

[0088] Further, the reward value of the decision tendency represents the degree of advantage or disadvantage of the decision tendency, and the higher the reward value, the greater the possibility that the decision tendency is determined as the current decision tendency.

[0089] In specific implementation, the state parameter is input into the first model, so that the reward value of the state parameter in the plurality of preset decision tendencies can be obtained, and further, the decision tendency with the maximum reward value is determined as the current decision tendency.

[0090] For example, in an embodiment, the input state parameter of the first model is [0.13, 0.21, 0.15, 0.22, 0.31, 0.45, 0.13, 0.61, 0.36], and the determined current decision tendency is 0.

[0091] S502, the second model determines the reward value of the state parameter corresponding to each scheduling rule in the plurality of scheduling rules based on the state parameter and the current decision tendency, and takes the scheduling rule with the maximum reward value as the current scheduling rule.

[0092] Specifically, the reward value of each scheduling rule represents the degree of advantage or disadvantage of the scheduling rule, and the higher the reward value, the greater the possibility that the scheduling rule is determined as the current scheduling rule. The current scheduling rule is used to indicate the scheduling group corresponding to the determination of the current scheduling operation according to the scheduling rule.

[0093] In specific implementation, the state parameter and the current decision tendency are input into the second model, so that the reward value of each scheduling rule corresponding to the state parameter can be obtained, and further, the scheduling rule with the maximum reward value is determined as the current scheduling rule.

[0094] For example, in combination with the above example, when the input state parameter is [0.13, 0.21, 0.15, 0.22, 0.31, 0.45, 0.13, 0.61, 0.36] and the determined current decision tendency is 0, [0.13, 0.21, 0.15, 0.22, 0.31, 0.45, 0.13, 0.61, 0.36, 0] is input into the second model in this step, and the second model outputs 1, that is, the determined current scheduling rule is FIFO.

[0095] S304, determining a scheduling group corresponding to the current scheduling operation according to the current scheduling rule; wherein the scheduling group is a combination of a workpiece, a machine and a worker.

[0096] Specifically, the scheduling group includes the workpiece, the machine and the worker, which indicates which worker processes which workpiece on which machine.

[0097] In a specific implementation, according to the current scheduling rule, the next workpiece to be processed is determined (based on the priority, deadline or other constraint conditions of the workpiece), and the machine available for processing the workpiece is determined (based on the availability, skill requirement or other factors of the machine), further, a worker is assigned to operate the selected machine to complete the processing of the workpiece (selected based on the skill level of the worker), and the combination of the workpiece, the machine and the worker is determined as the scheduling group corresponding to the current scheduling operation.

[0098] For example, in an embodiment, the scheduling group corresponding to the current scheduling operation is determined as [1, 2, 5], which represents that the worker 5 processes the workpiece 1 on the machine 2.

[0099] S305, decoding the scheduling group to a scheduling Gantt chart using a greedy decoding method.

[0100] Specifically, the earliest processable time of the workpiece, the earliest processable time of the machine and the earliest processable time of the worker are obtained, and the latest time among the three processable times is selected as the start processing time of the scheduling group. Further, the completion processing time of the scheduling group is determined according to the total processing time calculation formula in the first objective function, and the decoding of the scheduling group is completed. FIG. 6 is an implementation principle diagram of the greedy decoding according to an exemplary embodiment of the present application. The specific implementation principle and implementation process of the greedy decoding method can be referred to the description in the related art, which will not be described here.

[0101] For example, in an embodiment, combined with the above example, the completion processing time of the workpiece 1 in the last step is 15.0, the completion processing time of the machine 2 in the last step is 13.0, and the completion processing time of the worker 5 in the last step is 18.0, then the start time of the scheduling group is 18.0. Further, if the calculated total processing time is 8.0, then the decoded scheduling group is: the worker 5 processes the workpiece 1 on the machine 2, and the processing task starts at 18.0 and ends at 26.0.

[0102] Combined with FIG. 2 and the above description, for each scheduling operation, the scheduling solution is solved according to FIG. 2 until all scheduling operations are executed, and finally the output scheduling Gantt chart is the final obtained scheduling scheme.

[0103] It should be noted that Table 1 is a comparison table of the solving time of the scheduling method provided in an exemplary embodiment of the present application and the traditional meta-heuristic algorithm represented by ACO and NSGA-II.

[0104] Table 1

[0105] As can be seen from Table 1, the method provided in the present application can realize real-time scheduling solving, greatly improving the solving speed of the scheduling solution.

[0106] The embodiment provides a method for determining a current scheduling rule of a scheduling network, and the current scheduling rule can be determined based on input state parameters in a network prediction manner through the method, so that a scheduling group is determined based on the current scheduling rule, and a workshop scheduling is controlled.

[0107] The method provided in the embodiment has the following advantages. First, the workshop scheduling model and the production scheduling example are constructed, and the production scheduling example is solved through the workshop scheduling model, so that the workshop scheduling is realized, the effective production resources are reasonably arranged, the production process is optimized, and the production efficiency of the workshop is improved. Second, the skill level, fatigue degree and system carbon emission of workers are considered when the workshop scheduling model is constructed, the carbon emission generated in production is reduced under the condition of maintaining the production fairness of workers, a more scientific and sustainable workshop scheduling model can be established for a manufacturing enterprise, the production efficiency is improved, the carbon emission is reduced, and the sustainable development of the enterprise is of great significance. Third, the production scheduling example is solved through different search strategies, the generalization of the scheduling solution is improved, the production benefit of the workshop is improved, real-time scheduling solving is realized, the solving speed of the scheduling solution is greatly improved, and the scheduling speed is improved.

[0108] FIG. 7 is a flowchart of a third embodiment of the workshop scheduling method considering the skill level and fatigue degree of workers provided in the present application. FIG. 8 is an implementation principle diagram of training a scheduling network according to an exemplary embodiment of the present application. Please refer to FIGS. 7 and 8 simultaneously. The method provided in the embodiment has the following advantages. On the basis of the above-mentioned embodiment, the training process of the scheduling network includes:

[0109] S701, constructing a learning example by using the workshop scheduling model.

[0110] Specifically, in combination with the above description, a learning example can be constructed by using a workshop scheduling model according to production order data, worker data and workshop equipment data. The production order data includes workpiece name, process number, process set, process selectable machine set, process selectable machine processing basic time, workpiece material, workpiece quality, workpiece yield, and material utilization rate. The worker data includes worker name, worker skill set, worker first operation time, learning rate, fatigue accumulation rate, and fatigue recovery rate. The workshop equipment data includes machine name, man-machine interaction type, idle power, load power, auxiliary processing instrument set, processing instrument life, consumable set, unit time consumable amount, and machine position.

[0111] In a specific implementation, the production order data, the worker data and the workshop equipment data calculated in real time are input into the workshop scheduling model, so that the learning example is obtained.

[0112] S702, for each scheduling operation, when the number of scheduling operations does not reach a preset learning number, a scheduling solution corresponding to the current scheduling operation is calculated based on the current scheduling network by using the scheduling algorithm.

[0113] Specifically, the preset learning number is set according to actual needs, and in this embodiment, the specific value of the preset learning number is not limited. For example, in an embodiment, the preset learning number is 1000. The scheduling algorithm is selected according to actual needs, and in this embodiment, it is not limited. For example, in an embodiment, the scheduling algorithm is a genetic algorithm.

[0114] In a specific implementation, for each scheduling operation, the state parameters of the workshop at the current time are determined based on the production scheduling example and the scheduling Gantt chart at the current time. Further, the search strategy used for the current scheduling operation is determined, and then the current scheduling rule is determined based on the search strategy (when the search strategy is random search, a decision tendency is randomly selected from a plurality of preset decision tendencies as a current decision tendency, and a scheduling rule is randomly selected from a plurality of preset scheduling rules as a current scheduling rule; when the search strategy is network prediction, the state parameters of the workshop at the current time are input into the scheduling network, the scheduling network determines the reward values of each decision tendency in the plurality of preset decision tendencies corresponding to the state parameters, and the decision tendency with the maximum reward value is taken as the current decision tendency, the scheduling network determines the reward values of each scheduling rule in the plurality of preset scheduling rules corresponding to the state parameters, and the scheduling rule with the maximum reward value is taken as the current scheduling rule). Finally, the scheduling solution corresponding to the current scheduling operation is determined according to the current scheduling rule.

[0115] S703, the reward value of the scheduling solution is calculated according to the reward value calculation method.

[0116] Specifically, the reward value represents the degree of goodness of the scheduling solution, the higher the reward value, the better the scheduling solution, and the better the production efficiency it brings in the workshop scheduling.

[0117] Further, in combination with the above description, since the objective function of the workshop scheduling model includes a first objective function for minimizing the total production time, a second objective function for minimizing the system carbon emission, and a third objective function for minimizing the worker fatigue standard deviation. Therefore, the reward value includes three aspects of rewards, which are the reward based on the total production time, the reward based on the system carbon emission, and the reward based on the worker fatigue standard deviation.

[0118] In a specific implementation, the reward value of the scheduling solution is calculated according to the following formula:

[0119] Wherein, the r y is the reward value of the scheduling solution; the is the reward coefficient corresponding to the time-based reward of the yth scheduling operation; the is the time-based reward; the

[0120] is the reward coefficient corresponding to the carbon emission-based reward of the yth scheduling operation; the is the carbon emission-based reward; the is the reward coefficient corresponding to the human-based reward of the yth scheduling operation; and the is the human-based reward of the yth scheduling operation.

[0121] It should be noted that the and the will change according to the decision tendency of the yth scheduling operation. When the decision tendency is biased towards minimizing the total production time, the is set to 0.6, and the and the are set to 0.2. When the decision tendency is biased towards minimizing the system carbon emission, the is set to 0.6, and the and the are set to 0.2. When the decision tendency is biased towards minimizing the worker fatigue standard deviation, the is set to 0.6, and the and the are set to 0.2.

[0122] For example, in an embodiment, if the is 0.6, the is 0.5, the is 0.2, and the is 0.2.is 1.0, is 0.2, is -1.0, the reward value of the calculated scheduling solution is 0.3.

[0123] S704, the state parameters corresponding to the previous scheduling operation, the current state parameters, the current decision tendency, the current scheduling rule and the current calculated reward value are stored in the experience pool as a piece of scheduling experience.

[0124] In specific implementation, the state parameters corresponding to the previous scheduling operation, the current state parameters, the current decision tendency, the current scheduling rule and the reward value calculated by the reward value calculation formula are stored in the experience pool as a piece of scheduling experience.

[0125] For example, in an embodiment, in combination with the above example, when the state parameters corresponding to the previous scheduling operation are [0.13, 0.21, 0.15, 0.22, 0.31, 0.45, 0.13, 0.61, 0.36], the current state parameters are [0.15, 0.18, 0.16, 0.23, 0.29, 0.35, 0.15, 0.63, 0.39], the current decision tendency is 0, the current scheduling rule is 1 and the current calculated reward value is 0.3, in this step, the scheduling experience stored in the experience pool is [[0.13, 0.21, 0.15, 0.22, 0.31, 0.45, 0.13, 0.61, 0.36], [0.15, 0.18, 0.16, 0.23, 0.29, 0.35, 0.15, 0.63, 0.39], 0, 1, 0.3].

[0126] S705, when it is determined that model updating is needed, a preset number of pieces of scheduling experience are obtained from the experience pool.

[0127] Specifically, the preset number is set according to actual needs, and in this embodiment, the specific value of the preset number is not limited. For example, in an embodiment, the preset number is 16.

[0128] In specific implementation, every time 2 steps of scheduling operations are completed, it is determined that model updating is needed, at which time the preset number of pieces of scheduling experience are taken out from the experience pool. If only 1 step of scheduling operation is completed, it is determined that model updating is not needed, at which time the next step of scheduling operation is continued.

[0129] For example, in an embodiment, in combination with the above example, every time 2 steps of scheduling operations are completed, 16 pieces of scheduling experience are taken out from the experience pool.

[0130] S706, for each piece of scheduling experience, a predicted scheduling rule is predicted according to the current scheduling network to obtain predicted data.

[0131] Specifically, the prediction data includes a predicted decision tendency, a reward value of the decision tendency, a predicted scheduling rule, and a reward value of the scheduling rule.

[0132] Specifically, in combination with the above description, for each scheduling experience (including the state parameter corresponding to the previous scheduling operation, the current state parameter, the current decision tendency, the current scheduling rule, and the current calculated reward value), the current state parameter is input into the current scheduling network to obtain the predicted decision tendency, the reward value of the decision tendency, the predicted scheduling rule, and the reward value of the scheduling rule.

[0133] S707, evaluate the prediction loss according to the prediction data and the actual data of all scheduling experiences, and update the current scheduling network.

[0134] Specifically, the prediction loss is evaluated using the MES loss function on the prediction data (including the predicted decision tendency, the reward value of the decision tendency, the predicted scheduling rule, and the reward value of the scheduling rule) of all scheduling experiences and the actual data (the current decision tendency, the current scheduling rule, and the current calculated reward value) of all scheduling experiences. The scheduling network is updated according to the evaluation result.

[0135] Specifically, referring to FIG. 8, the first model and the second model are both composed of an evaluation network and a target network, the evaluation network of the first model is TDQN-E, and the target network is TDQN-T, the evaluation network of the second model is ADQN-E, and the target network is ADQN-T. When updating the scheduling network, the evaluation network in the scheduling network, i.e., TDQN-E and ADQN-E, can be updated.

[0136] Further, it is judged whether the target network update rule is met, i.e., the weight of the evaluation network is copied to the target network every 200 scheduling operations, i.e., the weight of TDQN-E is copied to TDQN-T, and the weight of ADQN-E is copied to ADQN-T.

[0137] S708, the scheduling solution corresponding to the current scheduling operation is calculated again until the number of scheduling operations reaches the preset learning number.

[0138] Specifically, the scheduling solution corresponding to the current scheduling operation is calculated again based on the method in step S702 until the number of scheduling operations reaches the preset learning number.

[0139] For example, in combination with the above example, the scheduling solution corresponding to the current scheduling operation is calculated again in this step until the number of scheduling operations reaches 1000 times.

[0140] Specifically, in combination with the foregoing description, after 1000 complete scheduling solutions are completed, learning is completed, and the scheduling network after training is completed is saved.

[0141] The embodiment provides a training method of a scheduling network. The scheduling network trained by the method can obtain a final trained scheduling network by continuously evaluating a prediction loss of predicted data and actual data and updating, and can solve a production scheduling example based on the scheduling network, improve generalization of a scheduling solution, improve production efficiency of a workshop, realize real-time scheduling solution, and greatly improve a solution speed of the scheduling solution.

[0142] Corresponding to the aforementioned embodiment of the workshop scheduling method considering the skill level and fatigue degree of workers, the application also provides an embodiment of a workshop scheduling device considering the skill level and fatigue degree of workers.

[0143] The embodiment of the workshop scheduling device considering the skill level and fatigue degree of workers provided by the application can be applied to a workshop scheduling equipment considering the skill level and fatigue degree of workers. The device embodiment can be realized by software, or realized by hardware or a combination of software and hardware. Taking software realization as an example, as a logical device, the device is formed by reading corresponding computer program instructions in a non-volatile memory to a memory and running by a processor of the workshop scheduling equipment where the device is located. From the hardware level, as shown in FIG. 9, it is a hardware structure diagram of the workshop scheduling equipment considering the skill level and fatigue degree of workers where the workshop scheduling device considering the skill level and fatigue degree of workers provided by the application is located. In addition to the processor, the memory, the network interface and the non-volatile memory shown in FIG. 9, the workshop scheduling equipment where the device is located in the embodiment can also include other hardware according to the actual function of the workshop scheduling device, and details are not described herein.

[0144] FIG. 10 is a structural schematic diagram of the embodiment one of the workshop scheduling device considering the skill level and fatigue degree of workers provided by the application. Please refer to FIG. 10, the device provided by the embodiment includes a construction module 1010 and a solution module 1020; wherein,

[0145] The construction module 1010 is configured to construct a workshop scheduling model, and construct a production scheduling example based on the workshop scheduling model according to production order data, worker data and workshop equipment data; wherein the workshop scheduling model is a workshop scheduling model considering the skill level and fatigue degree of workers;

[0146] The solution module 1020 is configured to solve the production scheduling example by using a scheduling algorithm, to obtain a scheduling solution of the production scheduling example;

[0147] The solving the production scheduling example by using the scheduling algorithm includes:

[0148] For each scheduling operation, based on the production scheduling example and the scheduling Gantt chart at the current time, the state parameters of the workshop at the current time are determined.

[0149] A search strategy used in the current scheduling operation is determined.

[0150] When the search strategy is random search, a scheduling rule is randomly selected from a plurality of preset scheduling rules as a current scheduling rule; when the search strategy is network prediction, the state parameters of the workshop at the current time are input into a scheduling network, so that the scheduling network determines reward values of the state parameters corresponding to each of the plurality of scheduling rules, and a scheduling rule with the largest reward value is selected as the current scheduling rule.

[0151] A scheduling group corresponding to the current scheduling operation is determined according to the current scheduling rule; wherein the scheduling group is a combination of workpieces, machines and workers.

[0152] The scheduling group is decoded to the scheduling Gantt chart using a greedy decoding method.

[0153] The device provided in the embodiment, in a first aspect, by constructing a workshop scheduling model and a production scheduling example, and solving the production scheduling example through the workshop scheduling model, realizes workshop scheduling, can reasonably arrange effective production resources, optimizes production processes, and improves the production efficiency of the workshop; in a second aspect, when constructing the workshop scheduling model, the skill level, fatigue degree and system carbon emission of workers are considered, in the case of maintaining the production fairness of workers, the carbon emission generated in production is considered to be reduced, which can help manufacturing enterprises to establish a more scientific and sustainable workshop scheduling model, improve the production efficiency, and reduce the carbon emission, which is of great significance for the sustainable development of enterprises; in a third aspect, by using different search strategies to solve the production scheduling example, the generalization of the scheduling solution is improved, the production benefit of the workshop is improved, real-time scheduling solution is realized, the solving speed of the scheduling solution is greatly improved, and the scheduling speed is improved.

[0154] The device of the embodiment can be used to execute the steps of the method embodiment shown in FIG. 1, and the specific implementation principles and implementation processes are similar, which will not be described here.

[0155] Please continue to refer to FIG. 9, the application also provides a workshop scheduling device considering the skill level and fatigue degree of workers, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to realize the steps of any one of the methods provided in the first aspect of the application.

[0156] The application also provides a computer readable storage medium having a computer program stored thereon, wherein the program is executed by a processor to realize the steps of any one of the methods provided in the application.

[0157] The implementation process of the functions and roles of each unit in the above device is specifically described in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0158] For the device embodiment, since it basically corresponds to the method embodiment, the relevant part can be seen from the part of the method embodiment. The device embodiment described above is only schematic, and the units shown as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or also distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the scheme of the present application according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0159] The above only describes the preferred embodiments of the present application and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the scope of protection of the present application.

Claims

1. A job shop scheduling method that takes into account skill levels and fatigue levels of workers, characterized by, The workshop scheduling method comprises: constructing a workshop scheduling model, and constructing a production scheduling instance based on the workshop scheduling model according to production order data, worker data and workshop equipment data; wherein the workshop scheduling model is a workshop scheduling model considering skill levels and fatigue degrees of workers; solving the production scheduling instance by using a scheduling algorithm to obtain a scheduling solution of the production scheduling instance; wherein solving the production scheduling instance by using a scheduling algorithm comprises: for each scheduling operation, determining state parameters of the workshop at the current time based on the production scheduling instance and a scheduling Gantt chart at the current time; determining a search strategy used for the current scheduling operation; when the search strategy is random search, randomly selecting a scheduling rule from a plurality of preset scheduling rules as a current scheduling rule; when the search strategy is network prediction, inputting the state parameters of the workshop at the current time into a scheduling network to determine, by the scheduling network, reward values of the state parameters corresponding to each scheduling rule in the plurality of scheduling rules, and taking a scheduling rule with the largest reward value as the current scheduling rule; determining a scheduling group corresponding to the current scheduling operation according to the current scheduling rule; wherein the scheduling group is a combination of a workpiece, a machine and a worker; decoding the scheduling group to a scheduling Gantt chart by using a greedy decoding mode; the scheduling network comprises a first model and a second model, and the scheduling network determines reward values of the state parameters corresponding to each scheduling rule in the plurality of scheduling rules, and takes a scheduling rule with the largest reward value as the current scheduling rule, which comprises: the first model determines reward values of the state parameters corresponding to a plurality of preset decision tendencies, and determines a decision tendency with the largest reward value as a current decision tendency; the second model determines reward values of the state parameters corresponding to each scheduling rule in the plurality of scheduling rules based on the state parameters and the current decision tendency, and takes a scheduling rule with the largest reward value as the current scheduling rule.

2. The method of claim 1, wherein the training process of the scheduling network comprises: constructing a learning instance by using the workshop scheduling model; for each scheduling operation, when the number of scheduling operations does not reach a preset learning number, calculating a scheduling solution corresponding to the current scheduling operation by using the scheduling algorithm based on the current scheduling network; calculating a reward value of the scheduling solution according to a reward value calculation method; storing state parameters of a previous scheduling operation, the current state parameters, a current decision tendency, a current scheduling rule and the current calculated reward value as a piece of scheduling experience in an experience pool; when it is determined that model updating is needed, obtaining a preset number of pieces of scheduling experience from the experience pool; for each piece of scheduling experience, predicting a scheduling rule based on the current scheduling network to obtain prediction data; evaluating prediction loss based on the prediction data and actual data of all pieces of scheduling experience, and updating the current scheduling network; calculating the scheduling solution corresponding to the current scheduling operation again until the number of scheduling operations reaches the preset learning number.

3. The method of claim 2, wherein, the calculation of the reward value of the scheduling solution comprises: The reward value of the scheduling solution is calculated according to the following formula: Wherein, the r y is the reward value of the scheduling solution; the corresponding reward coefficient of a time-based reward corresponding to a scheduling operation of the yth step; the is a time-based reward; the corresponding to the scheduling operation of the yth step; the for carbon emissions-based rewards; the corresponding reward coefficient of the reward corresponding to the scheduling operation of the yth step; the A reward based on human cause corresponding to the scheduling operation of the yth step is scheduled.

4. The method of claim 1, wherein, The preset plurality of scheduling rules include: {FIFO}, {FOFI}, {LPT}, {SPT}, {MCO}, {LTH}, {LCR, LCE, LTW}, {LCR, LCE, SOT}, {LCR, SMT, LTW}, {LCR, SMT, SOT}, {HCR, LCE, LTW}, {HCR, LCE, SOT}, {HCR, SMT, LTW}, {HCR, SMT, SOT}, {HCR, SMT, LTW}, {HCR, SMT, SOT}; The FIFO indicates selecting a workpiece-machine-worker scheduling group that can be processed earliest. The FOFI indicates selecting a workpiece-machine-worker scheduling group that can be completed earliest. The LPT indicates selecting a workpiece-machine-worker scheduling group with the longest processing time. The SPT indicates selecting a workpiece-machine-worker scheduling group with the shortest processing time. The MCO indicates selecting a workpiece-machine-worker scheduling group that increases the least emissions. The LTH indicates selecting a workpiece-machine-worker scheduling group that makes the least workers accumulate the most fatigue. The LCR indicates selecting a workpiece with the smallest completion rate. The HCR indicates selecting a workpiece with the largest completion rate. The LCE indicates selecting a machine with the smallest carbon emissions. The SMT indicates selecting a machine with the shortest processing time. The LTW indicates selecting a worker with the smallest current fatigue. The SOT indicates selecting a worker with the shortest operation time.

5. The method of claim 1, wherein, The workshop scheduling model is as follows: In the production system, there are n workpieces forming a workpiece task set J = (J1, J2, ..., Jn). n A set W = (W1, W2, ..., Wn) consisting of w workers. w In a machine set M = (M1, M2, ..., Mn) containing m machines... m Processing is performed on the workpiece J; each workpiece J i r with a fixed order i OP process i =(O i1 O i2 ,...,O ir Process O ij This indicates that the j-th process of workpiece i can be processed by selecting a machine from a specific subset of machine set M, and each worker W a Having a specific set of skills Worker skill refers to a worker's ability to operate specific machines, and each worker has a corresponding set of skill levels. When the equipment type is automated processing equipment, the skill level is quantified as worker W. a Operating machine M p The initial operation time; when the equipment is a human-machine collaborative processing equipment, the skill level is quantified as worker W. a First time operating machine M p The percentage of processing steps exceeding standard working hours; in each processing assignment, only when workpiece J... i Machine M p and worker W a Processing tasks can only be carried out when both are ready.

6. The method of claim 5, wherein, The objective function of the workshop scheduling model includes a first objective function for minimizing the total production time, a second objective function for minimizing system carbon emissions, and a third objective function for minimizing the standard deviation of worker fatigue; wherein, The first objective function is: wherein the T E is the total production time; the total processing time for the jth process of the workpiece i; said for the skill level of the worker operating the machine; the H is the incompressibility factor; the for the worker W a operating the machine M p times; said For worker W a learning rate; the the basic processing time of the jth operation of the workpiece i on the machine M p ; the MK m is the machine type; The second objective function is: f2 = min C E C E = C AE + C CM + C MR + C TR + C FP wherein said C E is the total carbon emissions of the system; said C AE is the total carbon emissions of the auxiliary processing equipment losses; said C CM is the total carbon emissions due to the use of processing consumables; said C MR is the total carbon emissions of the equipment operation of the manufacturing system to complete the manufacturing task; said C TR is the total carbon emissions of the handling during the processing; said C FP is the total carbon emissions of the workshop public equipment; The third objective function is: wherein the VF E worker fatigue standard deviation; the for the workers W a cumulative fatigue value at time t+1; the w is the number of workers in the production system; the R is a decision variable for the worker's work state; the R t R is a work load coefficient defined according to the type of machine operated by the worker; the For worker W a cumulative fatigue value at time t; the a fatigue accumulation rate determined by the physical condition of the worker; and The fatigue recovery rate is determined by the physical fitness of the worker. The constraint condition of the workshop scheduling model is: ensuring that only one machine processes each process of each workpiece; ensuring that only one worker processes each process of each workpiece; ensuring that at least one worker uses each machine; ensuring that a worker only processes the machines that the worker can operate; and ensuring that the start processing time of a subsequent process of each workpiece cannot be earlier than the end processing time of a previous process.

7. A plant scheduling device that takes into account skill levels and fatigue levels of workers, characterized by, The device includes a construction module and a solving module; wherein, The construction module is configured to construct a workshop scheduling model, and construct a production scheduling example based on the workshop scheduling model according to production order data, worker data, and workshop equipment data; wherein, the workshop scheduling model is a workshop scheduling model that considers the skill level and fatigue degree of a worker. The solving module is configured to solve the production scheduling example by using a scheduling algorithm to obtain a scheduling solution of the production scheduling example. The solving of the production scheduling example by using the scheduling algorithm includes: For each scheduling operation, based on the production scheduling example and the current state parameters of the scheduling Gantt chart at the current time, the state parameters of the workshop at the current time are determined. A search strategy used for the current scheduling operation is determined. A reward based on human cause corresponding to the scheduling operation of the yth step is scheduled. The preset plurality of scheduling rules include: {FIFO}, {FOFI}, {LPT}, {SPT}, {MCO}, {LTH}, {LCR, LCE, LTW}, {LCR, LCE, SOT}, {LCR, SMT, LTW}, {LCR, SMT, SOT}, {HCR, LCE, LTW}, {HCR, LCE, SOT}, {HCR, SMT, LTW}, {HCR, SMT, SOT}, {HCR, SMT, LTW}, {HCR, SMT, SOT}; The FIFO indicates selecting a workpiece-machine-worker scheduling group that can be processed earliest. The FOFI indicates selecting a workpiece-machine-worker scheduling group that can be completed earliest. The LPT indicates selecting a workpiece-machine-worker scheduling group with the longest processing time. The SPT indicates selecting a workpiece-machine-worker scheduling group with the shortest processing time. The MCO indicates selecting a workpiece-machine-worker scheduling group that increases the least emissions. The LTH indicates selecting a workpiece-machine-worker scheduling group that makes the least workers accumulate the most fatigue. The LCR indicates selecting a workpiece with the smallest completion rate. The HCR indicates selecting a workpiece with the largest completion rate. The LCE indicates selecting a machine with the smallest carbon emissions. The SMT indicates selecting a machine with the shortest processing time. The LTW indicates selecting a worker with the smallest current fatigue. The SOT indicates selecting a worker with the shortest operation time. The objective function of the workshop scheduling model includes a first objective function for minimizing the total production time, a second objective function for minimizing system carbon emissions, and a third objective function for minimizing the standard deviation of worker fatigue; wherein, The second objective function is: The fatigue recovery rate is determined by the physical fitness of the worker. The constraint condition of the workshop scheduling model is: ensuring that only one machine processes each process of each workpiece; ensuring that only one worker processes each process of each workpiece; ensuring that at least one worker uses each machine; ensuring that a worker only processes the machines that the worker can operate; and ensuring that the start processing time of a subsequent process of each workpiece cannot be earlier than the end processing time of a previous process. The device includes a construction module and a solving module; wherein, The construction module, used for constructing a workshop scheduling model, and constructing a production scheduling example based on the workshop scheduling model according to production order data, worker data, and workshop equipment data; wherein, the workshop scheduling model is a workshop scheduling model that considers the skill level and fatigue degree of a worker; The solving module, used for solving the production scheduling example by using a scheduling algorithm to obtain a scheduling solution of the production scheduling example; The solving of the production scheduling example by using the scheduling algorithm includes: For each scheduling operation, based on the production scheduling example and the current state parameters of the scheduling Gantt chart at the current time, the state parameters of the workshop at the current time are determined. A search strategy used for the current scheduling operation is determined. When the search strategy is a random search, a scheduling rule is randomly selected from a plurality of preset scheduling rules as a current scheduling rule; when the search strategy is a network prediction, a current state parameter of the workshop is input into a scheduling network to determine, by the scheduling network, a reward value of the state parameter corresponding to each of the plurality of scheduling rules, and a scheduling rule with a maximum reward value is selected as the current scheduling rule; A scheduling group corresponding to the current scheduling operation is determined according to the current scheduling rule, wherein the scheduling group is a combination of workpieces, machines and workers; The scheduling group is decoded to a scheduling Gantt chart using a greedy decoding mode; The scheduling network includes a first model and a second model, the scheduling network determines the reward value of the state parameter corresponding to each of the plurality of scheduling rules, and the scheduling rule with the maximum reward value is selected as the current scheduling rule, including: The first model determines the reward value of the state parameter corresponding to a plurality of preset decision tendencies, and determines the decision tendency with the maximum reward value as the current decision tendency; The second model determines the reward value of the state parameter corresponding to each of the plurality of scheduling rules based on the state parameter and the current decision tendency, and selects the scheduling rule with the maximum reward value as the current scheduling rule.

8. A plant scheduling apparatus that takes into account skill levels and fatigue levels of workers, characterized by, A computer program product, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the method of any one of claims 1-6 when executing the program.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the steps of the method of any one of claims 1-6.

Images