CN111898908A - Production line scheduling system and method based on multiple wisdom bodies - Google Patents

Abstract

The invention belongs to the technical field of production scheduling, and discloses a multi-agent-based production line scheduling system and method. The production scheduling system includes a global management module, a controller agent and a sub-controller agent. The global management module is the highest-level manager in the production scheduling system and has the highest management authority; the controller agent is the core of the production scheduling system. It includes a resource module, a planning calculation module and a task management module. The resource module is used to store the resource information of the sub-controller agent; the planning calculation module decomposes the tasks and sends them to the task management module; the task management module arranges the decomposed tasks in priority order Sorted and then distributed to the sub-controller agents; the sub-controller agents are used to execute the tasks sent from the task management module. Through the present invention, the independent operation ability of the intelligent body is ensured, and the work efficiency and management ability of the dispatching system are improved.

Description

A multi-agent-based production line scheduling system and method

technical field

The invention belongs to the technical field of production scheduling, and more particularly, relates to a production line scheduling system and method based on multi-agents.

Background technique

With the continuous advancement of science and technology, concepts such as intelligent manufacturing, distributed computing, the Internet of Things, and artificial intelligence have emerged as the times require. People are constantly putting forward their own needs for commodities. Personalization and diversification have become more and more people's primary choice. , which accordingly puts forward higher requirements for manufacturers, mainly reflected in the following aspects: (1) The previous "small variety and large batch" production mode is no longer applicable, and replaced by small batches with multiple varieties or even Single-piece customized production; (2) The current scheduling system is mostly coordinated by various functional modules. Once a module is abnormal, the system cannot run normally; (3) It can quickly respond to emergencies and have effective solutions (4) In the process of processing and production, due to resource conflicts, equipment hardware constraints and other restrictions, conflicts will occur between equipment.

In recent years, Multi-Agent System (MAS), as a research hotspot in the field of artificial intelligence, has been widely used in the construction of large-scale complex systems. Multi-agent systems refer to multiple distributed and parallel working agents. A computing system that cooperates to complete certain tasks, a multi-agent system has a certain degree of autonomy, interactivity and intelligence. The problem is solved, which provides a new idea and mode for the scheduling system.

Ningbo Saifu Technology Co., Ltd. has applied for the patent of "real-time scheduling system for factory intelligent workshop" (patent number: 201610403522.5). The patented real-time scheduling system for factory intelligent workshop uses intelligent and information technology to control workshop production lines, logistics and transportation systems, and production control. It can effectively improve the production management level of the factory by the dispatching system, but the alarm system is more to detect the temperature, brightness, air quality, noise, etc. of the environment, and does not involve the failure of production equipment. The collection and analysis of information and the production rescheduling caused by faults, and the feedback on the addition and removal of workshop equipment is not timely enough, which affects the practicability of the system; CN201611100675.9 Multi-agent-based workshop autonomous scheduling system and method, the The feature of the patented autonomous scheduling system is to set up corresponding intelligent bodies for all workpieces, equipment and logistics tools, collect data in the production process, and alarm in time when a fault occurs, which can ensure the robustness and reliability of the system to a certain extent, but it does The diagnosis of faults is relatively simple, the influence of the types of faults is not fully considered, and only one factory in a fixed location is considered in production control, and the distributed manufacturing resources of the enterprise are not taken into account, so it is not a universal scheduling system; that is, The current multi-agent scheduling system has the following problems: (1) The functions of the agents are too single, so that the coupling between each agent is too high, the autonomy of each agent is not enough, and it does not have the ability to solve problems independently; (2) The conflict coordination scheme between the agents is too simple, and the idle resources and capabilities of the remaining agents are not fully considered, and the scheduling results obtained are not good enough; (3) The rescheduling ability for emergencies is not enough, so it is impossible to The rescheduling scheme that yields the response.

SUMMARY OF THE INVENTION

In view of the above defects or improvement needs of the prior art, the present invention provides a production line scheduling system and method based on multi-agents, wherein the agents are used to control the management, resources, equipment, monitoring and other modules involved in the production and processing process of the enterprise. The functions and structures are encapsulated to obtain agents with independent functions, forming a corresponding multi-agent system, which reduces the coupling of agents between systems, ensures the ability of the agents to operate independently, and improves the work efficiency and efficiency of the scheduling system. management ability.

In order to achieve the above object, according to one aspect of the present invention, a multi-agent-based production line scheduling system and method is provided. The production scheduling system includes a global management module, a controller agent and a sub-controller agent, wherein:

The global management module is the highest-level manager in the production scheduling system, has the highest management authority, and is used to input task information, including receiving external task orders, task changes and task cancellations. The controller agent is connected, and receives the task result fed back by the controller agent, and evaluates the task result;

The controller agent is the core of the production scheduling system, including a resource module, a planning calculation module and a task management module, wherein:

The resource module is used to store the resource information of the sub-controller agent, including the quantity, type and equipment status of raw materials, and the resource information can be queried and monitored; the planning calculation module is connected to the resource module, and the planning After the computing module accepts the task from the global management module, it decomposes the task according to the resource information in the sub-controller agent of the resource module, and sends the decomposed task to the task management module, The task management module accepts the decomposition tasks from the planning calculation module, sorts the accepted decomposition tasks in a priority order, and distributes them to the sub-controller agents;

The sub-controller agent is used to execute the task sent from the task management module.

Further preferably, the controller agent further includes a coordination module, and the coordination module is used to coordinate among a plurality of the sub-control agents when the resources of the sub-control agents are missing, so that the sub-control agents The resources of the body meet the needs of performing tasks.

Further preferably, the controller agent further includes a monitoring module for monitoring the resource information status set by each sub-control agent, and feeding back the device status to the resource module, which is based on the current sub-controller. The resource information of the agent is updated.

Further preferably, the controller agent also includes a sub-controller management module for registering the information of the sub-controller under each controller module, including the number, communication address and port, and whether the sub-controller module is available. .

Further preferably, the number of controller agents in the production scheduling system is one or more.

Further preferably, for a production scheduling system with multiple control agents, the coordination module is further configured to implement communication between different control agents, so that resources between different control agents can be used in coordination.

According to another aspect of the present invention, there is provided a method for production scheduling by the above-mentioned production line scheduling system, the method comprising the following steps:

The S1 global management module receives the task, which includes the task of the specified machine tool and the task of the unspecified machine tool. The global management module sends the received task to the task management module. After the task management module receives the task, according to the content of the task Update the task table, equipment table and resource table in the task management module;

S2 For the task of the specified machine tool, query whether the status of the corresponding machine tool in the resource module is available. If the machine tool is available, select the machine tool; The controller intelligent body processes the task of the specified machine tool;

S3 For the task of unspecified machine tool, delete the information of the selected machine tool in the resource table of the task management module, and according to the machine tool in the current resource table, the planning calculation module decomposes the task of the unspecified machine tool into various machine tools The achievable method of the combination is calculated, and the score of each combination is calculated according to the state, resources and time cost of the sub-control agent corresponding to each machine tool, and the achievable method with the highest score is selected as the final realization method. According to the final implementation manner, the tasks of the unspecified machine tools are decomposed, and the task management module sends the decomposed tasks to the corresponding sub-controller management modules, so as to realize the processing of the tasks of the unspecified machine tools.

In general, compared with the prior art, the above technical solutions conceived by the present invention have the following beneficial effects:

1. The present invention uses intelligence to encapsulate the functions and structures of modules such as management, resources, equipment, monitoring, etc. involved in the production and processing process of the enterprise, and obtains intelligences with independent functions one by one, forming a corresponding multi-intelligence system, It reduces the coupling of the agents between systems and ensures the ability of the agents to operate independently;

2. The present invention can quickly complete the task decomposition and allocation and calculation planning through the interaction and cooperation between the agents, which improves the calculation planning ability in the existing scheduling system, and at the same time, the agents can maintain the mutual coordination of resources, etc., which enhances the The continuous stability of the system;

3. The present invention performs hierarchical management on product resources and other information according to the product structure, and uses corresponding formalized data for storage, which can respond to additions, deletions, and revisions of product information in a timely manner, thereby improving the information management level of the system;

4. The present invention selects the corresponding rescheduling method according to the type of equipment failure and the impact caused by the failure, and redistributes the original scheduling scheme to meet the task requirements and ensure the adaptability and robustness of the system.

Description of drawings

1 is a schematic structural diagram of a scheduling system constructed according to a preferred embodiment of the present invention;

Fig. 2 is the structure diagram of the controller agent and its sub-controller agent constructed according to the preferred embodiment of the present invention;

Fig. 3 is the task dispatch flow chart of the controller agent and its sub-controller agent constructed according to the preferred embodiment of the present invention;

Fig. 4 is the data flow diagram that the task of the controller agent constructed according to the preferred embodiment of the present invention is issued;

Fig. 5 is the flow chart of the task decomposition of the controller agent constructed according to the preferred embodiment of the present invention;

FIG. 6 is a flowchart of coordination among peer controller agents constructed in accordance with a preferred embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

As shown in FIG. 1 , the present invention relates to a multi-agent-based production line scheduling system and method. The corresponding dynamic scheduling system is mainly composed of a global management module and controller agents at various levels, and the controller agents include a task management module. , resource management module, planning module, coordination module, sub-controller management module, monitoring module and coordination module.

As the virtual system administrator of the system, the global management module is the message processing station of all the agents. As the information input of the whole system, the request information of all the agents in the lower layer is timely fed back to the global management module, and the global management module makes a decision after the decision is made. Forwarding, including equipment failure information, scheduling and rescheduling request information, etc., the global management module has the highest authority in the system, and when other agents conflict with it, the global management module makes the final decision.

The task management module in the controller agent manages the tasks of the agent, including task resource and demand detection, task priority sorting, task delivery, task status update, and emergency scheduling such as emergency task delivery and task cancellation to ensure the accuracy of task information in the agent.

The resource management module represents the manufacturing resource structure of the controller agent, including resource information, equipment information and some key information. The resource information includes basic information such as the type, quantity, and location of raw materials. Work progress, etc., for the equipment in the controller agent to simulate the work progress of its corresponding equipment, and the key information corresponds to map location information or process information.

The controller management module is responsible for the task allocation and management of the sub-controller agents. When the tasks are issued, the tasks are allocated reasonably according to the resource conditions of the sub-controllers, and the planning results returned by the sub-controller agents are summarized and calculated to obtain The optimal scheduling result is obtained, and the coordination request returned by the sub-controller agent is accepted, and the coordinated agent or resource is allocated to meet the requirement.

The planning module of the controller agent encapsulates the logic method of agent scheduling, and can flexibly adjust the corresponding algorithm logic according to the system requirements. In this module, at least two methods are required, namely, the task planning method and the rescheduling method. The former is the task decomposition method under normal conditions, and the latter is the rescheduling method when the equipment fails or the order is changed. Method model, parameter setting, etc.

The coordination module of the controller agent is responsible for coordination with other agents. When its own resources are insufficient, it will send assistance requests to other controller agents at the same level. At the same time, when the coordination module receives the coordination information from other agents, In the case of meeting the resource requirements of its own task, it allows resource assistance requests from other agents, and at the same time feeds back the request results to the assistance agent and the upper controller agent.

The monitoring module of the controller intelligent body is responsible for collecting the reading data of the mechanical condition monitoring equipment, mechanical fault diagnosis instrument, etc., and obtaining the working progress information of the equipment and analyzing the fault information of the equipment.

In order to realize the flexibility and efficiency of the system, each agent needs to set up a communication management module, which specifies the communication method between agents. The main types of communication interfaces between each device and the agent are Redis, OPC UA, etc. The specific communication methods can be flexibly replaced, thus realizing the coordinated work mode between systems.

As shown in Figure 2, when the scheduling system receives a task, the following operations are performed:

Step1: Start the global management module. The management agent runs at the center of the entire scheduling system, which is equivalent to the central server. After initializing the communication module, listen for registrations from other agents. In this way, the subsequent intelligence bodies are started, and the work progress of each intelligence body and equipment is displayed to the system administrator through the interface.

Step2: After the global management module accepts the task, it firstly judges the validity of the task, judges whether the task is in the required format, and sends it to the lower-level agent after passing the judgment, and then goes to step 3, otherwise the return task cannot be executed and an error is returned information.

Step3: After the task management module of the controller agent receives the task, it checks the resources to the resource management module according to the resources and equipment required by the task. After the resource detection is satisfied, it sorts and issues the current tasks of the system according to the task priority. , go to step 4, otherwise the return task cannot be executed and an error message will be returned.

Step4: After the planning and calculation module of the controller agent receives the task, it decomposes the task according to the capabilities and resources required by the task, and sends several sub-tasks to the corresponding sub-controller agent for planning and calculation. After the planning results of the sub-controller agents, the planning module performs summary calculations to obtain the optimal results, which are fed back to the global management module for final decision-making.

Step 5: After the global management module receives the planning result, it makes a final decision-making judgment. After the decision-making judgment is passed, the task is issued and executed.

As shown in Figures 3 and 4, the planning process of the controller agent and the sub-controller agent in the above step 4 is further described:

Step4.1: After the controller agent task management module receives the task, it determines whether the task specifies a device. If a device is specified, the controller agent queries the status of the specified device in the resource library according to the specified device. If the status is available, Select the device and go to step 4.2; if the status is not available, query the status of the device with the same or similar capabilities, select the device with available status, and go to step 4.2;

Step4.2: Add the selected device to the list of devices issued by the task, and delete the resources owned by the selected device from the task resource requirement list;

Step4.3: The planning calculation module decomposes the task according to the remaining resources and the equipment capabilities required by the task, decomposes the task into sub-task combinations of agents with different abilities, and stores them in the task list issued by the agent. Tasks are completed by agents (groups) with the same or similar abilities, and resources or constraints can be coordinated among each agent (groups) to resolve conflicts in the process.

Step4.4: For each subtask, the controller agent will assign weights according to the resource conditions of different agents and the cost of completing the task, so as to ensure that each subtask can obtain a task assignment plan with the smallest time cost. For resource conditions that cannot be met by an agent, it will coordinate with other agents to call resources to complete subtasks. Store the assigned agent combination in the agent sequence corresponding to each subtask.

Step 4.5: The decomposed subtasks are distributed to the corresponding sub-controller agents according to their assigned agent sequences to plan and calculate the specific implementation scheme, and wait for the planning results returned by them.

Step4.6: After the planning calculation module receives the planning scheme and coordination scheme of each sub-controller agent, it performs a certain weight score comparison, coordinates the possible conflicts between the agent schemes, and combines the overlapping operations to obtain The final overall plan and feedback to wait for final confirmation.

As shown in Figure 5, the task decomposition process of the planning calculation module in the controller agent is as follows:

Step1: After the planning calculation module receives the task, it determines whether the equipment required by the task belongs to the same type of controller agent group, if so, go to step 3; if not, go to step 2.

Step 2: Decompose the tasks according to the capabilities of the controller's agents, decompose the tasks into small tasks of different sets of agents, and send the small tasks to the corresponding groups of agents. For groups of agents with completely different abilities, they are allocated according to their abilities. For groups of agents with covered abilities, they should be allocated according to the efficiency of completing tasks, and priority should be given to groups with high efficiency.

Step3: If the sub-task can be completed by a single sub-controller agent, the sub-task does not need to be decomposed, go to step 4, if the sub-task cannot be completed by a single sub-controller agent, according to the time consumption and resource cost of the sub-controller agent Score the weights, and assign tasks according to the optimal results of the weighted scores.

Step4: After the sub-controller agent receives the sub-task, it performs planning and calculation according to its own resources and scores the corresponding cost weight of the planned result to obtain the planning score a; when the shared resources of the sub-controller agent are not empty When , the agent plans and calculates the task according to its own resources and shared resources, and scores the corresponding weights of the obtained planning results to obtain the planning score b. After the agent obtains the planning score a (or a and b), the planning result is fed back to the controller agent, and the controller agent performs summary calculation.

Step 5: After the controller agent obtains the planning result, it performs the final summary calculation to obtain the optimal coordinated planning result, and feeds it back to the global management module for confirmation of the final result.

As shown in Figure 6, the task coordination process between the controller agents is as follows:

Step1: When the controller agent resource is lacking and needs assistance, it will first check whether the shared resource is empty. If it is empty, it will return the assistance information to the parent controller agent; if it is not empty, it will be sent to the controller agent to which the shared resource belongs. Send assistance information to the body and wait for the assistance result;

Step2: When other controller agents receive the assistance information, they determine whether the shared resource is needed. If the task in the agent does not need the resource, it will return the assistable information; if the task in the agent needs the resource, it will return the non-assistable information. .

Step3: The controller agent receives the assistance information returned by other agents. If it is assistable information, it uses the resource to perform planning calculations and returns the planning result; if it is unassistable information, it returns the assistance information to the parent controller agent .

Step4: The parent controller agent specifies the corresponding coordination strategy according to the received return information.

When the equipment fails, the rescheduling process performed by the scheduling system is as follows:

Step1: The intelligent body monitoring module of the controller returns to receive the monitoring data of the corresponding monitoring equipment, and analyzes the working status of the equipment. When the working efficiency of the equipment decreases or the operation is interrupted due to parts wear, aging, fracture, or abnormal temperature and air pressure, the monitoring module needs to The fault information is encapsulated, indicating the type of fault, the consequences caused by the fault, etc., and finally the message is sent to the parent controller agent.

Step2: After the parent controller agent receives the fault information fed back from the monitoring module, the exception handling module analyzes the fault, regenerates the rescheduled task and the agent set, and performs the rescheduling and decomposition of the task, and at the same time, the fault information is sent upwards. Feedback to the global management module to inform users;

Step 3: After the planning calculation module receives the rescheduling request, it receives the rescheduled tasks and the required resource capability requirements, obtains the simulated data of the faulty device, removes some completed tasks, and reassigns the remaining tasks.

Step 4: According to the fault information returned by the monitoring module and the estimated repair time, assign tasks to the faulty equipment to varying degrees. If the equipment has a major failure, the estimated repair time exceeds the completion time required by the task, and this equipment will not be considered during reassignment; if the equipment failure is relatively minor and the estimated repair time does not exceed the completion time required by the task, this equipment will be added to the available In the scheduled equipment, the estimated repair time is added to the original estimated processing time.

Step5: The parent controller agent re-schedules the task, sends it to the child controller agent for planning calculation, summarizes the planning results returned by each agent, obtains a new rescheduling scheme and feeds it back to the global management agent, and sends it for execution.

Step6: When the equipment re-enters the system after repairing the fault, the sub-controller intelligent body equipment information is updated, and the intelligent bodies coordinate again to obtain the optimized solution after the device update, which is fed back to the upper-level intelligent body, and executed after confirmation.

Those skilled in the art can easily understand that the above are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention, etc., All should be included within the protection scope of the present invention.

Claims (7)

1. The utility model provides a production line dispatch system based on many wisdom, its characterized in that, this production dispatch system includes global management module, controller wisdom and sub-controller wisdom, wherein:

the global management module is a manager at the highest level in the production scheduling system, has the highest management authority, is used for inputting task information and comprises receiving external task orders, task changes and task revocation, and is connected with the controller intelligence, receives task results fed back by the controller intelligence and evaluates the task results;

the controller intelligence is the core of the production scheduling system, including a resource module, a planning calculation module and a task management module, wherein:

the resource module is used for storing resource information of the intelligent entity of the sub-controller, wherein the resource information comprises the quantity, the type and the equipment state of raw materials and is used for inquiring and monitoring; the planning computing module is connected with the resource module, decomposes the tasks according to resource information in a sub-controller intelligence in the resource module after receiving the tasks from the global management module, and sends the decomposed tasks to the task management module, and the task management module receives the decomposed tasks from the planning computing module, sorts the received decomposed tasks according to a priority order, and then distributes the decomposed tasks to the sub-controller intelligence;

and the sub-controller intelligence is used for executing the tasks sent by the task management module.

2. A multi-intelligence based production line scheduling system as recited in claim 1 wherein the controller intelligence further comprises a coordination module for coordinating among a plurality of the child control intelligence when the resources of the child control intelligence are missing so that the resources of the child control intelligence meet the requirements for performing tasks.

3. The method as claimed in claim 1, wherein the controller agent further comprises a monitoring module for monitoring the status of the resource information set by each sub-controller agent and feeding back the status of the equipment to the resource module, and the resource module is updated according to the resource information of the current sub-controller agent.

4. A multi-intelligence based production line scheduling system as claimed in claim 1 wherein said controller intelligence further comprises a sub-controller management module for registering information of sub-controllers under each controller module including number, communication address and port, and whether the sub-controller module is available.

5. A multi-intelligence based production line scheduling system as claimed in claim 2 wherein the number of controller intelligence in the production scheduling system is one or more.

6. A multi-agent based production line scheduling system according to claim 5 wherein, for a production scheduling system with multiple control agents, the coordination module is further configured to implement communication between different control agents so that resources between different control agents are used in coordination.

7. A method for production scheduling by a production line scheduling system according to any one of claims 1 to 6, comprising the steps of:

s1, the global management module receives tasks including the task of the appointed machine tool and the task of the unspecified machine tool, the global management module sends the received tasks to the task management module, and after the task management module receives the tasks, the task management module updates a task table, an equipment table and a resource table in the task management module according to the content in the tasks;

s2, for the task of the appointed machine tool, inquiring whether the state of the corresponding machine tool in the resource module is available, if the machine tool is available, selecting the machine tool, if not, selecting the available machine tool with the same or similar capacity, and adopting the sub-controller intelligence body where the selected machine tool is located to process the task of the appointed machine tool;

s3, for the task of the unspecified machine tool, the selected machine tool information is deleted in the resource table of the task management module, the planning calculation module decomposes the task of the unspecified machine tool into the realizable modes of various machine tool combinations according to the machine tool in the current resource table, the score of each combination is calculated according to the state of the sub-control intelligence body corresponding to each machine tool, the resource and the time cost for completing the task, the realizable mode with the highest score is selected as the final realization mode, the task of the unspecified machine tool is decomposed according to the final realization mode, and the task management module issues the decomposed task to the corresponding sub-controller management module, so that the task of the unspecified machine tool is processed.

Images