A flexible job shop energy-saving scheduling method based on multi-agent architecture

Through the multi-agent architecture, real-time scheduling and resource negotiation are achieved in the flexible operation workshop, the problems of energy waste and production disturbances in the manufacturing industry are solved, production efficiency and energy efficiency are improved, and system stability and response capabilities are improved.

CN115509188BActive Publication Date: 2024-11-01SOUTHWEST JIAOTONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211149236.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-09-21
Publication Date
2024-11-01
Estimated Expiration
2042-09-21

AI Technical Summary

Technical Problem

There are energy waste problems in the manufacturing industry, leading to global warming and energy crisis, and it is difficult to achieve efficient matching and adaptive scheduling of resources in the process of single-piece small-scale production and customized production, resulting in low production efficiency and energy efficiency.

Method used

Adopting a flexible work workshop energy-saving scheduling method based on multi-agent architecture, through real-time scheduling of distributed service units and multi-agent negotiation mechanism, the workpiece process allocation and equipment load can be optimized, load balancing and energy consumption can be minimized, and system stability can be improved. and response speed.

Benefits of technology

Real-time scheduling and allocation when new tasks arrive or disturbances occur, improves the stability and energy efficiency of the production process, reduces maximum completion time and energy consumption, and improves system robustness and responsiveness.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115509188B_ABST
    Figure CN115509188B_ABST
Patent Text Reader

Abstract

The present invention discloses a flexible job shop energy-saving scheduling method based on a multi-agent architecture. First, a configuration method for a workshop combined service unit is proposed, and based on this, workshop resources are encapsulated into corresponding agents to form a workshop multi-agent structure; secondly, a real-time scheduling optimization model under a double-layer agent architecture covering workpiece completion time, equipment load, and production energy consumption targets is established, and a multi-agent negotiation mechanism based on an improved multi-customer bargaining game is used to obtain the optimal scheduling solution and realize stable and efficient energy-saving production in the workshop manufacturing process. The present invention greatly improves the response time of real-time scheduling, ensures efficient and stable production of the production process, and can bring greater benefits to enterprises.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of real-time energy-saving scheduling of flexible job shops, and in particular relates to an energy-saving scheduling method for flexible job shops based on a multi-agent architecture. Background Art

[0002] With the development of social economy, energy consumption has become an issue that cannot be ignored in social development. At present, the manufacturing industry accounts for more than half of the total global energy consumption, and has doubled in the past 60 years. The waste of energy in the manufacturing industry is one of the main causes of global warming and energy crisis. Therefore, it is urgent to study the technology and methods of energy conservation and emission reduction in the manufacturing process, improve energy efficiency, and realize clean production in the manufacturing industry. Discrete manufacturing industry generally has problems such as single-piece small batch production and customized production, and production disturbances such as machine failures are often encountered in the actual production process, which brings great challenges to the production organization of the workshop. It is mainly reflected in three aspects: 1) How to effectively organize / describe the real-time status of production resources to achieve rapid matching of available resources considering energy saving and different production process requirements. 2) How to realize the monitoring and scheduling of resource conflicts / resource anomalies under multi-variety parallel production to achieve efficient and energy-saving production. 3) How to improve the response time of the manufacturing system when disturbances occur, so as to adapt to the rapid adaptive self-adjustment and self-energy saving when emergencies occur. Summary of the invention

[0003] In view of the above problems, the present invention provides a flexible job shop energy-saving scheduling method based on a multi-agent architecture.

[0004] A flexible job shop energy-saving scheduling method based on a multi-agent architecture of the present invention comprises the following steps:

[0005] Step 1: Description and assumptions of the flexible job shop energy-saving scheduling problem.

[0006] The real-time scheduling problem based on distributed service units is: there are u service units in a workshop, and each service unit has m u Each workpiece process can only be performed on one machine in a service unit.

[0007] Conditional assumptions:

[0008] (1) The transfer time of workpieces between different devices is not considered.

[0009] (2) Unless a machine failure occurs, once a workpiece is processed, it cannot be interrupted. The unfinished process after a machine failure needs to be reprocessed, and there is a tight constraint between the previous and next processes of the same task.

[0010] (3) A workpiece can only be served in one service unit at a time.

[0011] (4) Processing time consists of two parts: processing time and preparation time.

[0012] (5) The cache capacity between service units is infinite.

[0013] Parameter definition:

[0014] n is the number of workpieces, U is the number of service units, m u is the number of devices in service unit u, is the kth device in service unit u, O ij is the jth process of the i-th workpiece, For process O ij The processing completion time in service unit u, is the load of machine k in service unit u, For Machine The load power, For Machine The no-load power, For process O ij In the machine Upper processing, W u is the load of service unit u, For process O ij In the machine The start processing time on For process O ij In the machine Processing time on the uk For Machine The idle time, For process O ij In the machine Tool change time, δ i is the comprehensive benefit of player i, δ b It is the optimal profit of the system during the game.

[0015] Step 2: Establish a mathematical model for energy-saving scheduling based on service units.

[0016] At the workshop level, when a new workpiece arrives, the service units negotiate with each other based on the equipment status, and assign different processes of the workpiece to different service units. Under the premise of ensuring load balancing, the maximum completion time of the workshop is minimized, thereby ensuring the production efficiency of the workshop; inside the service unit, each service unit produces according to its own plan; when a new task arrives in the workshop, the service unit receives the workpiece information from the workshop and the status information from the equipment to make decisions and generate corresponding production plans; in order to improve production efficiency and energy efficiency, the multi-agent scheduling system based on the service unit considers the following three optimization objectives:

[0017] (1) Minimize the maximum completion time

[0018]

[0019] (2) Minimize the total equipment load in the service unit while minimizing the workpiece completion time to achieve equipment load balancing:

[0020]

[0021] Where W t Represents the total load of all equipment in the workshop. Under the premise of minimizing the total completion time, the optimization goal is to minimize the total load of all service units, thereby achieving load balancing of the service units.

[0022] (3) Minimize the production energy consumption of all service units:

[0023]

[0024] The production energy consumption of a workshop is the sum of the production energy consumption of the equipment in the service unit, and the production energy consumption of the equipment is the sum of the no-load energy consumption and the load energy consumption.

[0025] Constraints:

[0026] An artifact can only be served in one service unit at a time, and can only be served on one device:

[0027]

[0028] Two adjacent processes of the same workpiece have a close-before-close-after constraint:

[0029]

[0030] Step 3: Combination service unit configuration mechanism based on processing capabilities.

[0031] In a flexible discrete manufacturing workshop, flexibility refers to the flexibility of the processing equipment. Flexible processing equipment can process product processes in a variety of modes. From the perspective of completing a process or a group of processes, there are two types:

[0032] (1) A process or a group of processes can only be processed by one device, that is, only one device has the service capacity for this process.

[0033] (2) A process is processed on one of multiple optional equipment.

[0034] For type (1), a single device can independently complete a corresponding process and there is no alternative equipment, so the service unit built with this device as the core is the final service unit.

[0035] As for type (2), the service unit devices in the same area need further discussion during configuration, which can be divided into the following forms:

[0036] a. The equipment model type is exactly the same or the process service capabilities are exactly the same.

[0037] b. The equipment types are different, but they have the same process service capabilities.

[0038] c. The equipment types are inconsistent. While having the same process service capabilities, some of the equipment also have other different process service capabilities.

[0039] For cases a and b, equipment with exactly the same process service capabilities can be configured as a combined service unit, while for the last case, the equipment needs to be distinguished separately, that is, the equipment in each service unit after configuration should have the same process service capabilities; after the above configuration is completed, a distance check is required, that is, equipment with a long distance inside the service unit needs to be configured as an independent service unit: if the minimum transfer distance between the equipment inside the service unit and other equipment is greater than the distance from the equipment to the next service unit, if the latter is greater than the former, then the equipment needs to be configured as a new service unit. If the above problem still exists after reconfiguration, then the above steps are repeated.

[0040] Step 4: Establish a multi-agent structure based on service units.

[0041] The multi-agent structure based on service units includes the following four agents:

[0042] 1) SA: Workshop Agent, which is the virtual body of the workshop-level controller; it is deployed in the workshop master server and is responsible for managing all service units in the workshop. After obtaining the workshop resource information, it registers and generates the corresponding resource Agent. When a new workpiece arrives, it activates the corresponding JA and allocates different processes in the JA to the corresponding UA.

[0043] 2) UA: Service unit Agent, that is, the virtual body of the service unit-level controller; deployed in the distributed terminals of the workshop, divided into the minimum service unit agent and the combined service unit agent; responsible for managing the subordinate sub-service units / equipment. When a new workpiece arrives, UA obtains JA-related information and its service unit resource information, negotiates with other UAs to select the corresponding process for processing, allocates tasks to related equipment within the service, and arranges the start time and completion time of each task.

[0044] 3) MA: Machine Agent, which is a virtual entity of a processor, executor or cache. It is responsible for saving and managing basic information, status information and other attribute information of manufacturing resources, so as to keep track of the real-time status and available status of manufacturing resources at any time. It has the function of negotiating with other equipment agents. When a task arrives, it can actively capture relevant information about workpiece processing and send processing time information to JA.

[0045] 4) JA: Artifact Agent, which is a virtual entity of the flow entity; it is used to capture all task information. Its attributes include the number of jobs, the processing time of each process on different equipment, the available processing time, and the completion time. When a new task arrives, the artifact agent captures the artifact information in a timely manner, registers in the SA, and communicates and negotiates with relevant intelligent agents.

[0046] Step 5: Energy-saving scheduling solution based on two-layer multi-agent bargaining game.

[0047] The real-time scheduling problem of the service unit multi-agent system can be regarded as an N-person game with complete information; the game between agents is defined as the players solving the profit distribution problem through negotiation; the utility function is the corresponding optimization objective function. To construct the game model between service unit agents, three elements need to be determined, which are described as:

[0048] G={F i ; S i ; δ i} (6)

[0049] Among them, F i It is a set of players participating in the game; here, there are two levels of game processes: workshop-service unit level and service unit-machine level: at the workshop level, UA is the bargaining player in the bargaining game process, and SA is the bargaining player; at the service unit level, MA is the bargaining player in the bargaining game process, and UA is the bargaining player; here, the bargaining players make actions in turn during the game process, and the choice of each participant will affect the choices of other participants.

[0050] S i is the strategy of the game process; at the workshop level, the strategy of each player UA corresponds to the unprocessed process in the optional unprocessed JA; at the service unit level, the strategy of each player corresponds to the first to-be-processed process of all JAs assigned to the service unit.

[0051] δ i For the corresponding benefits, different benefit functions correspond to different levels. Since the optimization goal is the minimum value of the corresponding objective function, the inverse of the objective function in the corresponding layer is the benefit of the players in the game.

[0052] Different types of intelligent agents communicate, interact and play games based on bargaining games, which mainly include bargaining games between service units, bargaining games within service units and Nash equilibrium of multi-customer bargaining games.

[0053] The specific process of bargaining game between service units is as follows:

[0054] S11: The multi-agent system obtains task-related information, activates the corresponding Agent in the JADE platform, generates the corresponding JA, MA, UA, and SA parses the processes contained in the JA.

[0055] S12: Each UA takes the optimization target as its own profit function, that is, the inverse of the objective function of each player is the profit of the corresponding player. The unprocessed process in JA is the strategy of the game. u UAs correspond to u bargaining parties in the game. The purpose of bargaining between each player is to maximize his own profit in the game process.

[0056] S13: All UAs make all possible strategies based on their own status information to form their own strategy set, form the solution with the greatest benefit to themselves and a set of acceptable solutions, and send them to the SA for the counter-offer process.

[0057] S14: SA combines all UAs in various feasible solutions and calculates the system's benefit value according to formula (1). The Nash equilibrium of the game process is obtained through bargaining.

[0058] S15: According to the Nash equilibrium result, each UA obtains a set of pending operations and adds it to the pending operation queue of the service unit.

[0059] S16: Repeat the above process until all processes in JA are assigned to corresponding UAs.

[0060] The specific process of bargaining game within the service unit is as follows:

[0061] S21: Inside the service unit, UA obtains task-related information. The m MAs correspond to the m bargaining parties in the game. The inverse of the optimization target is the profit function of the players. Each MA plays against each other to select the processing procedure with the highest profit for itself.

[0062] S22: The unscheduled processes assigned to UA form a set of strategies for tasks to be processed.

[0063] S23: Each player MA selects a processable process, forms a strategy set and calculates the corresponding benefits, proposes the expected optimal solution and the acceptable solution, and sends them to the UA.

[0064] S24: UA receives the strategy set sent by MA, integrates the solutions proposed by other players, and obtains the Nash equilibrium of the game process through bargaining; if there are multiple UAs with corresponding processing capabilities in the workshop, the UA enters the workshop level and bargains with other UAs to form a final solution.

[0065] S25: MA obtains information about the required processing steps based on the bargaining game results and updates the relevant attributes of MA and JA;

[0066] S26: Repeat the above steps until all processes of all workpieces are arranged and completed.

[0067] The Nash equilibrium of the multi-customer bargaining game is as follows:

[0068] For non-cooperative static games, Nash equilibrium is widely referred to as the point at which each player has no reason to change his or her behavior. Therefore, the Nash equilibrium point of the game process is the optimal value sought; the SPNE point is an N-tuple strategy. During the game process, any unilateral deviation of any participant cannot increase its expected return.

[0069] S represents the strategy set of player i in the game;

[0070]

[0071] V i Represents the corresponding profit function in the strategy set:

[0072]

[0073] Since the three optimization objectives are values ​​of different dimensions, the comprehensive benefit of player i in strategy k is:

[0074]

[0075] The optimal value of the game process is:

[0076]

[0077] The beneficial technical effects of the present invention are:

[0078] 1) The present invention designs a real-time scheduling framework, which can realize the real-time allocation of processing tasks when new tasks arrive or interference occurs, thereby eliminating the deviation from the original plan after the interference occurs, making the production process more stable and efficient.

[0079] 2) The present invention designs a distributed combined service unit configuration method, which can realize hierarchical processing of workshops when interference occurs. Compared with the existing methods, the system stability is greatly improved and the robustness is improved by more than 20%.

[0080] 3) This invention designs a multi-agent negotiation mechanism based on multi-client bargaining game, which can improve production efficiency and energy efficiency. Compared with the unimproved bargaining game method, the maximum completion time is reduced by more than 11% and the energy consumption is reduced by more than 10.8%. Compared with other existing methods, there are also improvements to varying degrees. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Figure 1 Configure a flow chart for the composite service unit;

[0082] Figure 2 It is the structure diagram of the workshop based on the combined service unit;

[0083] Figure 3 This is the structure diagram of the workshop multi-agent based on the combined service unit;

[0084] Figure 4 Gantt chart for scheduling results;

[0085] Figure 5 Gantt chart of scheduling results for scenario 1;

[0086] Figure 6 Gantt chart of scheduling results for scenario 2;

[0087] Figure 7 Gantt chart of scheduling results for scenario 3. DETAILED DESCRIPTION

[0088] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0089] The present invention is based on a multi-agent structure of a combined service unit configured based on processing capabilities and a multi-agent negotiation mechanism based on bargaining games. Manufacturing tasks are scheduled by the edge nodes of the service unit agents themselves, and the workshop agents act as central nodes to make monitoring decisions, thereby reducing the load on the central nodes, improving the robustness of the system and the flexibility of system resource configuration when dynamic interference occurs. By combining and configuring each service unit, the information flow, material flow, and control flow are combined into the service unit, and then with the help of distributed multi-agent technology, different production tasks, resources, and service units are modeled as agents with specific goals and attributes. Through the logical configuration between various elements, the manufacturing system is modeled as a multi-agent system based on service units. Different types of agents communicate, interact, and play games based on bargaining games to achieve the optimization of the production operation process. Specifically, the following steps are included:

[0090] Step 1: Description and assumptions of the flexible job shop energy-saving scheduling problem.

[0091] The real-time scheduling problem based on distributed service units is: there are u service units in a workshop, and each service unit has m u Each workpiece process can only be performed on one machine in a service unit. The main purpose of the flexible job shop scheduling problem based on service units is to assign each process to a suitable service unit and determine the best machine for each process based on real-time manufacturing information, so as to achieve optimization goals such as minimizing completion time and reducing workshop energy consumption. In order to solve this problem, the following assumptions are made:

[0092] (1) The transfer time of workpieces between different devices is not considered.

[0093] (2) Unless a machine failure occurs, once a workpiece is processed, it cannot be interrupted. The unfinished process after a machine failure needs to be reprocessed, and there is a tight constraint between the previous and next processes of the same task.

[0094] (3) A workpiece can only be served in one service unit at a time.

[0095] (4) Processing time consists of two parts: processing time and preparation time.

[0096] (5) The cache capacity between service units is infinite.

[0097] Parameter definition:

[0098] n is the number of workpieces, U is the number of service units, m u is the number of devices in service unit u, is the kth device in service unit u, O ij is the jth process of the i-th workpiece, For process O ij The processing completion time in service unit u, is the load of machine k in service unit u, For Machine The load power, For Machine The no-load power, For process O ij In the machine Upper processing, W u is the load of service unit u, For process O ij In the machine The start processing time on For process O ij In the machine Processing time on the uk For Machine The idle time, For process Oij In the machine Tool change time, δ i is the comprehensive benefit of player i, δ b It is the optimal profit of the system during the game.

[0099] Step 2: Establish a mathematical model for energy-saving scheduling based on service units.

[0100] At the workshop level, when a new workpiece arrives, the service units negotiate with each other based on the equipment status, and assign different processes of the workpiece to different service units. Under the premise of ensuring load balancing, the maximum completion time of the workshop is minimized, thereby ensuring the production efficiency of the workshop; inside the service unit, each service unit produces according to its own plan; when a new task arrives in the workshop, the service unit receives the workpiece information from the workshop and the status information from the equipment to make decisions and generate corresponding production plans; in order to improve production efficiency and energy efficiency, the multi-agent scheduling system based on the service unit considers the following three optimization objectives:

[0101] (1) Minimize the maximum completion time

[0102]

[0103] (2) Minimize the total equipment load in the service unit while minimizing the workpiece completion time to achieve equipment load balancing:

[0104]

[0105] Where W t Represents the total load of all equipment in the workshop. Under the premise of minimizing the total completion time, the optimization goal is to minimize the total load of all service units, thereby achieving load balancing of the service units.

[0106] (3) Minimize the production energy consumption of all service units:

[0107]

[0108] The production energy consumption of a workshop is the sum of the production energy consumption of the equipment in the service unit, and the production energy consumption of the equipment is the sum of the no-load energy consumption and the load energy consumption.

[0109] Constraints:

[0110] An artifact can only be served in one service unit at a time, and can only be served on one device:

[0111]

[0112] Two adjacent processes of the same workpiece have a close-before-close-after constraint:

[0113]

[0114] Step 3: Combination service unit configuration mechanism based on processing capabilities.

[0115] In a flexible discrete manufacturing workshop, flexibility refers to the flexibility of the processing equipment. Flexible processing equipment can process product processes in a variety of modes. From the perspective of completing a process or a group of processes, there are two types:

[0116] (1) A process or a group of processes can only be processed by one device, that is, only one device has the service capacity for this process.

[0117] (2) A process is processed on one of multiple optional equipment.

[0118] For type (1), a single device can independently complete a corresponding process and there is no alternative equipment, so the service unit built with this device as the core is the final service unit.

[0119] As for type (2), the service unit devices in the same area need further discussion during configuration, which can be divided into the following forms:

[0120] a. The equipment models and types are completely the same or the process service capabilities are completely the same, such as the same equipment clusters formed according to the clustering principle.

[0121] b. The equipment types are inconsistent, but have the same process service capabilities, such as a fully flexible job shop.

[0122] c. The equipment types are inconsistent. While having the same process service capabilities, some of the equipment also have other different process service capabilities, such as some finishing equipment also has rough processing functions.

[0123] For cases a and b, equipment with exactly the same process service capabilities can be configured as a combined service unit, while for the last case, the equipment needs to be distinguished separately, that is, the equipment in each service unit after configuration should have the same process service capabilities; after the above configuration is completed, a distance check is required, that is, equipment with a long distance inside the service unit needs to be configured as an independent service unit: if the minimum transfer distance between the equipment inside the service unit and other equipment is greater than the distance from the equipment to the next service unit, if the latter is greater than the former, then the equipment needs to be configured as a new service unit. If the above problem still exists after reconfiguration, then the above steps are repeated.

[0124] In a fully flexible workshop with a large workshop scale, there will be a large number of devices with the same process service capacity in the service unit. The reasonable configuration of the combined service unit will have different impacts on the different optimization goals of the production system. That is, if there are too many devices in the same service unit, the robustness and response time of the system will decrease. In order to ensure that the manufacturing system can maintain a high controllability and logistics transfer efficiency, the number of devices in the combined service unit should be 3-5, which can achieve higher optimization performance and stability of the system. The configuration process of the combined service unit is as follows: Figure 1 As shown, the structure of the workshop based on the combined service unit after configuration is as follows Figure 2 shown.

[0125] Step 4: Establish a multi-agent structure based on service units.

[0126] In traditional scheduling strategies, all tasks are centrally assigned to the corresponding machines by a distribution system. The decision model is centralized, and the machine does not interact with other resources, resulting in no consideration of the real-time status information of other resources. Therefore, due to unpredictable exceptions, there is often a certain error between the plan and its actual execution. The multi-agent structure based on the service unit of the present invention includes the following four agents:

[0127] 1) SA: Workshop Agent, which is the virtual body of the workshop-level controller; it is deployed in the workshop master server and is responsible for managing all service units in the workshop. After obtaining the workshop resource information, it registers and generates the corresponding resource Agent. When a new workpiece arrives, it activates the corresponding JA and allocates different processes in the JA to the corresponding UA.

[0128] 2) UA: Service unit Agent, that is, the virtual body of the service unit-level controller; deployed in the distributed terminals of the workshop, divided into the minimum service unit agent and the combined service unit agent; responsible for managing the subordinate sub-service units / equipment. When a new workpiece arrives, UA obtains JA-related information and its service unit resource information, negotiates with other UAs to select the corresponding process for processing, allocates tasks to related equipment within the service, and arranges the start time and completion time of each task.

[0129] 3) MA: Machine Agent, which is a virtual entity of a processor, executor or cache. It is responsible for saving and managing basic information, status information and other attribute information of manufacturing resources, so as to keep track of the real-time status and available status of manufacturing resources at any time. It has the function of negotiating with other equipment agents. When a task arrives, it can actively capture relevant information about workpiece processing and send processing time information to JA.

[0130] 4) JA: Artifact Agent, which is a virtual entity of the flow entity; it is used to capture all task information. Its attributes include the number of jobs, the processing time of each process on different equipment, the available processing time, and the completion time. When a new task arrives, the artifact agent captures the artifact information in a timely manner, registers in the SA, and communicates and negotiates with relevant intelligent agents.

[0131] Multi-agent systems based on service units such as Figure 3 As shown in the figure, agents of the same level bargain with each other and with the superior agent. When a new task is released to the workshop, the JA with relevant attributes is activated in the SA, and the JA releases the unprocessed process to form a queue of processes to be processed. At the same time, each service unit Agent negotiates and negotiates with each other according to its own available status, and the SA assigns different processes of different JAs to the corresponding UA according to the game results. The process constantly interacts with the UA. Inside the service unit, the MA negotiates with other MAs according to its own status, and finally assigns the process to the appropriate MA for processing. After the processing is completed, the agents update their respective information, and the JA continues to release the unprocessed process until all workpieces are processed. When a disturbance event occurs, if it is a disturbance such as a machine failure, the service unit first quickly reconfigures internally, and the UA decides which MAs can continue to process the process. When the disturbance is the arrival of a new order, the SA activates the corresponding JA and adds the relevant process to the queue to be processed, thereby realizing real-time scheduling of workshop production.

[0132] Step 5: Energy-saving scheduling solution based on two-layer multi-agent bargaining game.

[0133] The real-time scheduling problem of the service unit multi-agent system can be regarded as an N-person game with complete information; the game between agents is defined as the players solving the profit distribution problem through negotiation; the utility function is the corresponding optimization objective function. To construct the game model between service unit agents, three elements need to be determined, which are described as:

[0134] G={F i ; S i ; δ i} (6)

[0135] Among them, F i It is a set of players participating in the game; here, there are two levels of game processes: workshop-service unit level and service unit-machine level: at the workshop level, UA is the bargaining player in the bargaining game process, and SA is the bargaining player; at the service unit level, MA is the bargaining player in the bargaining game process, and UA is the bargaining player; here, the bargaining players make actions in turn during the game process, and the choice of each participant will affect the choices of other participants.

[0136] Si is the strategy of the game process; at the workshop level, the strategy of each player UA corresponds to the unprocessed process in the optional unprocessed JA; at the service unit level, the strategy of each player corresponds to the first to-be-processed process of all JAs assigned to the service unit.

[0137] δ i For the corresponding benefits, different benefit functions correspond to different levels. Since the optimization goal is the minimum value of the corresponding objective function, the inverse of the objective function in the corresponding layer is the benefit of the players in the game.

[0138] Different types of intelligent agents communicate, interact and play games based on bargaining games, which mainly include bargaining games between service units, bargaining games within service units and Nash equilibrium of multi-customer bargaining games.

[0139] Bargaining game between service units

[0140] When a new order arrives, SA activates the corresponding JA and allocates different processes in JA to appropriate UAs through the game between UAs. The specific process is as follows:

[0141] S11: The multi-agent system obtains task-related information, activates the corresponding Agent in the JADE platform, generates the corresponding JA, MA, UA, and SA parses the processes contained in the JA.

[0142] S12: Each UA takes the optimization target as its own profit function, that is, the inverse of the objective function of each player is the profit of the corresponding player. The unprocessed process in JA is the strategy of the game. u UAs correspond to u bargaining parties in the game. The purpose of bargaining between each player is to maximize his own profit in the game process.

[0143] S13: All UAs make all possible strategies based on their own status information to form their own strategy set, form the solution with the greatest benefit to themselves and a set of acceptable solutions, and send them to the SA for the counter-offer process.

[0144] S14: SA combines all UAs in various feasible solutions and calculates the system's benefit value according to formula (1). The Nash equilibrium of the game process is obtained through bargaining.

[0145] S15: According to the Nash equilibrium result, each UA obtains a set of pending operations and adds it to the pending operation queue of the service unit.

[0146] S16: Repeat the above process until all processes in JA are assigned to corresponding UAs.

[0147] At the shop floor level, if disturbances such as the arrival of new orders occur, the SA activates the corresponding JA, and the JA releases the corresponding process to enter the shop floor game process and allocates the JA to the corresponding UA.

[0148] Bargaining game within service unit

[0149] The UA layer is responsible for assigning the processes assigned by SA to the UA to the appropriate MA. The first process to be processed by each JA assigned to the UA forms a set of processes to be assigned. In each game process, the set of processes to be assigned serves as the strategy set of the players. The specific process is as follows:

[0150] S21: Inside the service unit, UA obtains task-related information. The m MAs correspond to the m bargaining parties in the game. The inverse of the optimization target is the profit function of the players. Each MA plays against each other to select the processing procedure with the highest profit for itself.

[0151] S22: The unscheduled processes assigned to UA form a set of strategies for tasks to be processed.

[0152] S23: Each player MA selects a processable process, forms a strategy set and calculates the corresponding benefits, proposes the expected optimal solution and the acceptable solution, and sends them to the UA.

[0153] S24: UA receives the strategy set sent by MA, integrates the solutions proposed by other players, and obtains the Nash equilibrium of the game process through bargaining; if there are multiple UAs with corresponding processing capabilities in the workshop, the UA enters the workshop level and bargains with other UAs to form a final solution.

[0154] S25: MA obtains information about the required processing steps based on the bargaining game results and updates the relevant attributes of MA and JA;

[0155] S26: Repeat the above steps until all processes of all workpieces are arranged and completed.

[0156] When abnormal events such as equipment failure occur, the corresponding MA attributes are changed accordingly, and the UA quickly organizes reconfiguration to promptly reduce the impact of the abnormal event. When the impact of the disturbance cannot be resolved within the service unit, the UA renegotiates with other UAs to promptly eliminate the impact of the corresponding disturbance.

[0157] The Nash equilibrium of the multi-customer bargaining game is as follows:

[0158] For non-cooperative static games, Nash equilibrium is widely referred to as the point at which each player has no reason to change his or her behavior. Therefore, the Nash equilibrium point of the game process is the optimal value sought; the SPNE point is an N-tuple strategy. During the game process, any unilateral deviation of any participant cannot increase its expected return.

[0159] S represents the strategy set of player i in the game;

[0160]

[0161] V i Represents the corresponding profit function in the strategy set:

[0162]

[0163] Since the three optimization objectives are values ​​of different dimensions, the comprehensive benefit of player i in strategy k is:

[0164]

[0165] The optimal value of the game process is:

[0166]

[0167] In order to find the equilibrium point of the game, the process of solving the equilibrium point of the multi-customer bargaining game proposed in the present invention is shown in Algorithm. Different from the traditional bargaining game, the present invention expands the game players to multiple customers, and in the process of merchants making strategies (each time the processing process is assigned to the processing machine), the game players on the customer side select their own optional strategies and calculate the corresponding profit values, then select the strategy with the maximum profit and alternative plans to form their own optional strategy set, that is, the bargaining acceptable plan, and make a bargaining process to the player UA. The game players on the merchant side finally integrate the plans provided by all customers and select the plan with the maximum profit, that is, the overall optimal strategy of the system as the bargaining plan. In this process, each customer wants to achieve the plan with the best profit for themselves, while the merchant side wants to maximize its own profit, that is, adopt the overall optimal plan of the system, thereby forming a bargaining game process.

[0168] The game process is as follows:

[0169]

[0170]

[0171] Example:

[0172] In order to demonstrate the effectiveness and superiority of the proposed method, a proof-of-concept prototype system was built on the JADE platform based on the jdk8.0 development environment to simulate the real scene in a production workshop based on service units. The real-time scheduling method based on multi-customer bargaining game was written in Java language and encapsulated into the corresponding agent. The prototype system was experimentally simulated and verified on a PC with Intel(R) Core(TM) i5@2.60GHz and 8GB RAM.

[0173] The present invention adopts the flexible job shop standard example proposed by Kacem for verification. The standard example library has been used by a large number of scholars in the verification process of workshop scheduling algorithms. In order to verify the effectiveness of the present invention in real-time energy-saving scheduling, the Kacem8×8 standard example extended data in the existing research is used for verification. The workpiece and machine energy consumption information is shown in Table 1.

[0174] Table 1 Workpiece process information

[0175]

[0176]

[0177] Each set of data corresponds to the machine processing preparation time, processing time and processing power. The Gantt chart of the scheduling result is as follows Figure 4 The comparison results with the existing multi-agent methods are shown in Table 2.

[0178] Table 3 Comparison of scheduling results

[0179]

[0180] It can be seen that the algorithm based on the multi-layer service unit structure proposed in the present invention has good superiority in solving multi-objective problems. It can be seen that in the study using the multi-agent method, the method proposed in the present invention can reduce the completion time by up to 11% and the total energy consumption by up to about 31%. Therefore, it can be proved that the method proposed in this article has good performance in multi-objective optimization of workshop scheduling, especially in energy consumption.

[0181] In order to verify the effectiveness of the scheduling scheme of the present invention when disturbance occurs, tests are carried out in three scenarios. The relevant scheduling Gantt charts are as follows: Figure 5-7 As shown in Table 3, the comparison results of the optimization objectives and existing related methods are shown in Table 3, and the comparison of robustness and response time is shown in Table 4.

[0182] Table 3 Comparison of scheduling results under different disturbance conditions

[0183]

[0184] Table 4 Robustness and response time comparison

[0185]

[0186] The specific situation of scenario 1 is: machine 2 fails at time 1 and is repaired at time 4, and machine 5 fails at time 11 and is repaired at time 13. The specific situation of scenario 2 is: workpiece 9 arrives at time 2. The specific situation of scenario 3 is: the above anomalies occur at the same time.

Claims

1. A flexible job shop energy-saving scheduling method based on a multi-agent architecture, characterized in that: The following steps are involved: Step 1: Description and assumptions of the energy-saving scheduling problem of flexible job shops; The real-time scheduling problem based on distributed service units is: there are u service units in a workshop, and each service unit has m u Each workpiece process can only be performed on one machine in a service unit; Conditional assumptions: (1) The transfer time of workpieces between different devices is not considered; (2) Unless a machine failure occurs, once a workpiece is processed, it cannot be interrupted. After a machine failure, the unfinished process needs to be reprocessed, and there is a close constraint between the previous and next processes of the same task; (3) A workpiece can only be served in one service unit at a time; (4) Processing time consists of two parts: processing time and preparation time; (5) The cache capacity between service units is infinite; Parameter definition: n is the number of workpieces, U is the number of service units, m u is the number of devices in service unit u, is the kth device in service unit u, O ij is the jth process of the i-th workpiece, For process O ij The processing completion time in service unit u, is the load of machine k in service unit u, For Machine The load power, For Machine The no-load power, For process O ij In the machine Upper processing, W u is the load of service unit u, For process O ij In the machine The start processing time on For process O ij In the machine Processing time on the uk For Machine The idle time, For process O ij In the machine Tool change time, δ i is the comprehensive benefit of player i, δ b The best benefit of the system during the game; Step 2: Establish a mathematical model for energy-saving scheduling based on service units; At the workshop level, when a new workpiece arrives, the service units negotiate with each other based on the equipment status, and assign different processes of the workpiece to different service units. Under the premise of ensuring load balancing, the maximum completion time of the workshop is minimized, thereby ensuring the production efficiency of the workshop; inside the service unit, each service unit produces according to its own plan; when a new task arrives in the workshop, the service unit receives the workpiece information from the workshop and the status information from the equipment to make decisions and generate corresponding production plans; in order to improve production efficiency and energy efficiency, the multi-agent scheduling system based on the service unit considers the following three optimization objectives: (1) Minimize the maximum completion time (2) Minimize the total equipment load in the service unit while minimizing the workpiece completion time to achieve equipment load balancing: Where W t Represents the total load of all equipment in the workshop. Under the premise of minimizing the total completion time, the optimization goal is to minimize the total load of all service units, so as to achieve load balancing of service units; (3) Minimize the production energy consumption of all service units: The production energy consumption of the workshop is the sum of the production energy consumption of the equipment in the service unit, and the production energy consumption of the equipment is the sum of the no-load energy consumption and the load energy consumption; Constraints: An artifact can only be served in one service unit at a time, and can only be served on one device: Two adjacent processes of the same workpiece have a close-before-close-after constraint: Step 3: Combination service unit configuration mechanism based on processing capability; In a flexible discrete manufacturing workshop, flexibility refers to the flexibility of the processing equipment. Flexible processing equipment can process product processes in a variety of modes. From the perspective of completing a process or a group of processes, there are two types: (1) A process or a group of processes can only be processed by one device, that is, only one device has the service capacity for this process; (2) One process is processed on one of multiple optional equipments; For type (1), a single device can independently complete a corresponding process and there is no alternative device, so the service unit built with this device as the core is the final service unit; As for type (2), the service unit devices in the same area need further discussion during configuration, which can be divided into the following forms: a. The equipment model and type are completely the same or the process service capabilities are completely the same; b. The equipment types are different, but they have the same process service capabilities; c. The equipment types are inconsistent. While they have the same process service capabilities, some of them also have other different process service capabilities; For cases a and b, equipment with exactly the same process service capabilities can be configured as a combined service unit, while for the last case, the equipment needs to be distinguished separately, that is, the equipment in each service unit after configuration should have the same process service capabilities; after the above configuration is completed, a distance test is required, that is, the equipment with a long distance inside the service unit needs to be configured as an independent service unit: if the minimum transfer distance between the equipment inside the service unit and other equipment is greater than the distance from the equipment to the next service unit, if the latter is greater than the former, then the equipment needs to be configured as a new service unit. If the above problem still exists after reconfiguration, the above steps are repeated; Step 4: Establish a multi-agent structure based on service units; The multi-agent structure based on service units includes the following four agents: 1) SA: Workshop Agent, which is the virtual body of the workshop-level controller. It is deployed in the workshop master server and is responsible for managing all service units in the workshop. After obtaining workshop resource information, it registers and generates the corresponding resource Agent. When a new workpiece arrives, it activates the corresponding JA and allocates different processes in the JA to the corresponding UA. 2) UA: Service unit Agent, i.e., the virtual body of the service unit-level controller; deployed in the distributed terminals of the workshop, divided into the minimum service unit agent and the combined service unit agent; responsible for managing the subordinate service units / equipment. When a new workpiece arrives, the UA obtains JA-related information and its service unit resource information, negotiates with other UAs to select the corresponding process for processing, allocates tasks to relevant equipment within the service, and arranges the start time and completion time of each task; 3) MA: Machine Agent, i.e., a virtual entity of a processor, executor, or cache; responsible for storing and managing basic information, status information, and attribute information of manufacturing resources, so as to understand the real-time status and availability of manufacturing resources at any time; It has the function of negotiating with other equipment agents. When a task arrives, it can actively capture the relevant information of workpiece processing and send processing time information to JA. 4) JA: Artifact Agent, which is a virtual entity of the flow entity; used to capture all task information, its attributes include the number of jobs, the processing time of each process on different equipment, the processing time, and the completion time; when a new task arrives, the artifact agent captures the artifact information in time, registers in the SA, and communicates and negotiates with the relevant intelligent agents; Step 5: Energy-saving scheduling solution based on two-layer multi-agent bargaining game; The real-time scheduling problem of the service unit multi-agent system can be regarded as an N-person game with complete information; the game between agents is defined as the players solving the profit distribution problem through negotiation; the utility function is the corresponding optimization objective function. To construct the game model between service unit agents, three elements need to be determined, which are described as: G={F i ;S i ;d i } (6) Among them, F i It is a set of players who participate in the game. Here, there are two levels of game processes: workshop-service unit level and service unit-machine level. At the workshop level, UA is the bargaining player in the bargaining game, and SA is the bargaining player. At the service unit level, MA is the bargaining player in the bargaining game, and UA is the bargaining player. Here, the bargaining players make actions in turn during the game, and the choice of each player will affect the choices of other players. S i is the strategy of the game process; at the workshop level, the strategy of each player UA corresponds to the unprocessed process in the optional unprocessed JA; at the service unit level, the strategy of each player corresponds to the first to-be-processed process of all JAs assigned to the service unit; δ i For the corresponding benefits, different benefit functions correspond to different levels. Since the optimization goal is the minimum value of the corresponding objective function, the inverse of the objective function in the corresponding layer is the benefit of the players in the game. Different types of intelligent agents communicate, interact and play games based on bargaining games, including bargaining games between service units, bargaining games within service units and Nash equilibrium of multi-customer bargaining games.

2. According to the flexible job shop energy-saving scheduling method based on multi-agent architecture according to claim 1, it is characterized in that: The specific process of the bargaining game between the service units is as follows: S11: The multi-agent system obtains task-related information, activates the corresponding Agent in the JADE platform, generates the corresponding JA, MA, and UA, and the SA parses the processes contained in the JA; S12: Each UA takes the optimization goal as its own benefit function, that is, the inverse of the objective function of each player is the benefit of the corresponding player. The unprocessed process in JA is the strategy of the game. u UAs correspond to u bargaining parties in the game. The purpose of bargaining between each player is to maximize their own benefits in the game process. S13: All UAs make all possible strategies based on their own status information to form their own strategy set, form the solution with the greatest benefit to themselves and the set of acceptable solutions, and send them to the SA for the counter-offer process; S14: SA combines all UAs in various feasible solutions, calculates the system's benefit value according to formula (1), and obtains the Nash equilibrium of the game process through bargaining; S15: According to the Nash equilibrium result, each UA obtains a set of pending operations and adds it to the pending operations queue of the service unit; S16: Repeat the above process until all processes in JA are assigned to corresponding UAs.

3. According to the flexible job shop energy-saving scheduling method based on multi-agent architecture according to claim 1, it is characterized in that: The specific process of the bargaining game within the service unit is as follows: S21: Inside the service unit, the UA obtains task-related information. The m MAs correspond to the m bargaining parties in the game. The inverse of the optimization target is the profit function of the parties. Each MA plays against each other to select the processing procedure with the highest profit for itself. S22: The unscheduled processes assigned to the UA form a set of strategies for tasks to be processed; S23: Each player MA selects a processable process, forms a strategy set and calculates the corresponding benefits, proposes the expected optimal solution and the acceptable solution, and sends them to the UA; S24: UA receives the strategy set sent by MA, integrates the solutions proposed by other players, and obtains the Nash equilibrium of the game process through bargaining; if there are multiple UAs with corresponding processing capabilities in the workshop, the UA enters the workshop level and bargains with other UAs to form a final solution; S25: MA obtains information about the required processing steps based on the bargaining game results and updates the relevant attributes of MA and JA; S26: Repeat the above steps until all processes of all workpieces are arranged and completed.

4. According to the flexible job shop energy-saving scheduling method based on multi-agent architecture according to claim 1, it is characterized in that: The Nash equilibrium of the multi-customer bargaining game is specifically: For non-cooperative static games, Nash equilibrium becomes the point where each player has no reason to change his behavior. Therefore, the Nash equilibrium point of the game process is the optimal value sought. The Nash equilibrium point of the game is an N-tuple strategy. During the game process, any unilateral deviation of any participant cannot increase its expected return. S represents the strategy set of player i in the game; V i Represents the corresponding profit function in the strategy set: Since the three optimization objectives are values ​​of different dimensions, the comprehensive benefit of player i in strategy k is: The optimal value of the game process is: