Dynamic scheduling method for digital twin workshop lightweight information interaction

Abstract

The invention discloses a dynamic scheduling method for digital twin workshop lightweight information interaction. According to the method, five function modules including an equipment state module, a workpiece state module, a scheduling-associated environment state module, a high-response dynamic production scheduling module and a disturbance identification and dynamic scheduling module are included. And through interaction between the intelligent agent in the high-response dynamic production scheduling module and the scheduling-associated environment state module, training is carried out to obtain a dynamic scheduling intelligent agent. And when an abnormal event occurs in the production process, the disturbance identification and dynamic scheduling module immediately obtains the current equipment state module and the current workpiece state module, calls the trained dynamic scheduling intelligent agent, outputs a lightweight scheduling result consisting of a workpiece JSON data block and an equipment JSON data block, and displays the lightweight scheduling result. The method is used for simulation and decision making of production scheduling of a digital twin workshop platform, and high efficiency of dynamic scheduling calculation of a production workshop and light weight of scheduling information transmission are achieved.

Description

Technical Field

[0001] The present invention relates to production scheduling technology of manufacturing process, designs a dynamic scheduling method for lightweight information interaction of digital twin workshops, and belongs to the field of advanced manufacturing. Background Art

[0002] Dynamic scheduling is an upgrade of the production workshop scheduling optimization problem. Due to the large number of uncertainties in the production process and the frequent dynamic insertion of orders, the normal operation of the workshop will have a great impact. Once an abnormal event occurs in the production workshop, it is usually necessary to stop work for abnormal processing and then recalculate the new production schedule. In this process, equipment and personnel are idle, resulting in resource waste, affecting the delivery period, bringing delivery delay costs, etc., which greatly affects production efficiency and increases manufacturing costs. For this type of problem, on the one hand, how to improve the efficiency of scheduling calculations, and on the other hand, how to quickly obtain production status.

[0003] In production scheduling calculations, conventional meta-heuristic algorithms represented by genetic algorithms are only suitable for scheduling optimization calculations when there is sufficient time before production due to their long calculation time. Stopping processing for a long time during the production process to reschedule calculations will waste a lot of time in the workshop. The reinforcement learning algorithm with multiple rule combinations can solve this problem well. The intelligent agent is trained before the start of processing. When encountering disturbance problems such as inserted orders and machine failures during the processing, the trained intelligent agent can be used to make quick decisions on the current problem, provide new scheduling solutions in a short time, and minimize time loss.

[0004] In terms of obtaining production status, the current development of digital twin workshop technology has provided a monitoring platform foundation for obtaining production status. However, the information involved in production is complex and huge, and the high collection frequency leads to a large data interaction load. Therefore, it is of great significance to realize lightweight information interaction and transmission of status information such as workpieces, equipment, and production environment related to scheduling, which is of great significance to improving the dynamic scheduling response capability of production. Summary of the invention

[0005] The present invention proposes a dynamic scheduling method for lightweight information interaction of a digital twin workshop, which is used to solve the rapid adjustment of production plans and information interaction after uncertain production abnormal events occur during the production process, thereby improving production efficiency.

[0006] A dynamic scheduling method for lightweight information interaction in a digital twin workshop, comprising: an equipment status module, a workpiece status module, a scheduling-related environment status module, a high-response dynamic scheduling module, and a disturbance identification and dynamic scheduling module.

[0007] (1) Device status module

[0008] It is used to reflect the status of a single generated device. For multiple devices in a workshop, multiple device modules can be generated. The total number of devices is set to m (m ≥ 1), and the device index is set to j (j ∈ [1, m]). Each device module contains: device index j, the start time set of the processing operation assigned to the device The end time set of the processing operations assigned to the equipment Index of the workpiece assigned to the processing operation on the machine The set of devices M is represented as follows:

[0009]

[0010] (2) Workpiece status module

[0011] Used to reflect the status of a single workpiece. The total number of workpieces is set to n (n ≥ 1), and the workpiece index is set to i (i ∈ [1, n]). Each workpiece module contains: workpiece index i, the total number of workpiece processes The number of processes currently completed by the workpiece The start time of each process of the workpiece The end time of each process of the workpiece The equipment index set assigned to each process of processing the workpiece A set J of n workpieces n It is expressed as follows:

[0012]

[0013] (3) Scheduling related environment status module

[0014] It is used to obtain the workshop status in real time through the digital twin workshop platform, including 10 parameters at the three levels of equipment, workpiece, and processing for monitoring the production workshop status. The environmental state S set is represented as follows:

[0015] S=[s mu ,s mp ,s ju ,s jp ,s jn ,s jm ,s jr ,s jt ,s mt ] (1)

[0016]

[0017]

[0018] Average equipment utilization mu, is the average utilization rate of each equipment in the workshop at the current moment. The utilization rate of each equipment is the ratio of the working time of the equipment to the production time, as shown in formula (2).

[0019]

[0020] The average number of processes arranged for each machine tool s mp , which is used to indicate the load level of each device, is calculated by calculating the number of processes that each machine tool has processed and averaging them, as shown in formula (3).

[0021]

[0022] Average probability of workpiece placement s ju , the workpiece availability rate is calculated by counting the ratio of the processing time of each workpiece to the current maximum working time of each workpiece in the workshop, and finally the average workpiece availability rate is obtained by averaging, as shown in formula (4).

[0023]

[0024] Average process completion rate jp , which is used to count the process completion status of each workpiece, and average the completed process / total process of each workpiece. As shown in formula (5).

[0025]

[0026] 6 indicators of processing jn 、s jo 、s jm 、s jr 、s jt 、s mt ) is the status value of the workpiece and equipment under the current production task, which is directly obtained through data monitoring.

[0027] (4) Highly responsive dynamic production scheduling module

[0028] A multi-rule combination mode is adopted to realize high-response dynamic production scheduling calculation, among which the multi-rules include 12 types: giving priority to the workpiece with the least remaining processes (SOP), giving priority to the workpiece with the shortest processing time in the next process (SSO), giving priority to the workpiece with the short remaining processing time (SRPT), giving priority to the workpiece with the shortest total processing time (SRM), giving priority to the workpiece that completes the processing the fastest (SLP), giving priority to the workpiece with the lowest machine tool utilization (SMP), giving priority to the workpiece with the longest remaining processing time (LOP), giving priority to the workpiece with the long processing time in the next process (LSO), giving priority to the workpiece with the long remaining processing time (LRPT), giving priority to the workpiece with the longest total processing time (LRM), giving priority to the workpiece that completes the processing the slowest (LLP), and giving priority to the workpiece with the highest machine tool utilization (LMP).

[0029] Two decisions need to be made in the production scheduling calculation. Decision 1 is to determine the next workpiece process to be arranged. This decision is achieved by selecting one of the above 12 rules. On this basis, the agent is trained based on the reinforcement learning algorithm to achieve the selection of 12 rules. In reinforcement learning, the agent interacts and iterates with the environment state module. The interaction process includes: obtaining the environment state, making choices based on multiple rules, obtaining rewards under the choice, and updating the environment state. The agent's parameters are optimized through iteration. Among them, the reward function is based on the equipment utilization rate as the target. The difference between the utilization rate of each device before the multi-rule selection starts and the utilization rate of each device after the selection is observed. If the change in equipment utilization rate is positive, it means that the behavior is favorable, otherwise it means that the behavior is unfavorable, as shown in formula (6).

[0030] r t =mr t -mr t-1 (6)

[0031] Where r t For the reward of making a decision, mr t is the sum of all equipment utilization after the decision, mr t-1 It is the sum of all equipment utilization rates before the decision is made. The intelligent agent that has completed iterative optimization directly decides the selection of the process in the production scheduling calculation.

[0032] Decision 2 is to allocate processing equipment to the workpiece process. After the process is determined, the set of processing equipment that can process the process is determined. Among these processing equipment sets, if there are idle devices, idle devices are randomly selected for allocation. If there are no idle devices, the equipment that completes the current task the fastest is selected as the allocated equipment for the process.

[0033] (5) Disturbance identification and dynamic scheduling module

[0034] In workshop production, disturbances are divided into two categories: one is immediate disturbance, including equipment failure, time delay, etc., and the other is delayed disturbance, including order changes, emergency order inquiries, etc. After the disturbance occurs, the current equipment status module and the current workpiece status module are immediately obtained to determine the current completed process, the process being completed, and the process to be completed. The dynamic scheduling time is determined based on the immediate disturbance and delayed disturbance, and the intelligent agent is called to make a quick decision on the process to be completed in combination with the current equipment status, and the workpiece JSON data block and the equipment JSON data block are output.

[0035] The workpiece JSON data block contains 6 keys, i represents the workpiece index, num represents the total number of workpiece processes, cur represents the workpiece's currently completed process index, i_start represents the start time of each process of the workpiece, i_end represents the end time of each process of the workpiece, and mac represents the equipment index assigned to each process of the workpiece. The following is an example, there are two workpieces, workpiece 1 has 3 processes, the currently completed process is 0, the processing equipment index of the first process is 2, starting at time 0 and ending at time 3; the processing equipment index of the second process is 3, starting at time 5 and ending at time 10; the processing equipment index of the third process is 1, starting at time 10 and ending at time 16. [

[0037] {i:"0001",num:3,cur:0,i_start:"0,5,10",i_end:"3,10,16",mac:"2,3,1"}

[0038] {i:"0002",num:4,cur:0,i_start:"0,4,8,12",i_end:"4,8,12,15",mac:"1,2,2,3"}]

[0039] The equipment JSON data block contains 4 keys, j represents the equipment index, j_start represents the start time of all processes assigned to the equipment, j_end represents the end time of all processes assigned to the equipment, and job represents the workpiece index of all processes assigned to the equipment. In the following example, there are 3 equipments, and two processes are assigned to equipment 1. The first one belongs to workpiece index 2, starts at time 0 and ends at time 4; the second one belongs to workpiece index 1, starts at time 10 and ends at time 16. [

[0041] {j:"01",j_start:"0,10",j_end:"4,16",job:"2,1"}

[0042] {j:"02",j_start:"0,4,8",j_end:"3,8,12",job:"1,2,2,3"}

[0043] {j:"03",j_start:"12",j_end:"15",job:"2"} ]

[0045] The workpiece data block and equipment data block are output as lightweight files in JSON format, which can be directly used in the digital twin workshop system for dynamic scheduling simulation and decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0047] Figure 1 This is a schematic diagram of the implementation process of the dynamic scheduling method DETAILED DESCRIPTION

[0048] The present invention proposes a dynamic scheduling method for lightweight information interaction of digital twin workshops, which is used to solve the problem of rapid adjustment of production plans and information interaction after uncertain production events occur during the production process, thereby improving production efficiency. The present invention is further described below in conjunction with the accompanying drawings and specific implementation methods:

[0049] like Figure 1 As shown, a dynamic scheduling method for lightweight information interaction of digital twin workshops includes: equipment status module, workpiece status module, scheduling-related environment status module, high-response dynamic scheduling module, disturbance identification and dynamic scheduling module.

[0050] The equipment status module is used to reflect the status of a single generated equipment. For multiple equipment in the workshop, multiple equipment modules can be generated. The total number of equipment is set to m (m ≥ 1), and the equipment index is set to j (j ∈ [1, m]). Each equipment module contains: equipment index j, the start time set of the processing operation assigned to the equipment The end time set of the processing operations assigned to the equipment Index of the workpiece assigned to the processing operation on the machine The set of devices M is represented as follows:

[0051]

[0052] The workpiece status module is used to reflect the status of a single workpiece. The total number of workpieces is set to n (n ≥ 1), and the workpiece index is set to i (i ∈ [1, n]). Each workpiece module contains: workpiece index i, the total number of workpiece processes The number of processes currently completed by the workpiece The start time of each process of the workpiece The end time of each process of the workpiece The equipment index set assigned to each process of processing the workpiece A set J of n workpieces n It is expressed as follows:

[0053]

[0054] The scheduling-related environment status module is used to obtain the workshop status in real time through the digital twin workshop platform, including 10 parameters at three levels: equipment, workpiece, and processing, for monitoring the production workshop status. The set of environment status S is represented as follows:

[0055] S=[s mu ,s mp ,s ju ,s jp ,s jn ,s jm ,s jr ,s jt ,s mt ] (7)

[0056]

[0057] The equipment class status includes the average equipment utilization s mu , the average number of processes arranged for each machine tool s mp , which can be expressed as follows:

[0058]

[0059] The workpiece status includes the average probability of the workpiece s ju , average process completion rate s jp , which can be expressed as follows:

[0060]

[0061] 6 indicators of processing status (s jn 、s jo 、s jm 、s jr 、s jt 、s mt ) is the status value of the workpiece and equipment under the current production task, which is directly obtained through data monitoring.

[0062] The high-response dynamic production scheduling module includes 12 decision rules, reward functions, and intelligent agents, wherein the decision rules include: giving priority to the workpiece with the least remaining processes (SOP), giving priority to the workpiece with the shortest processing time in the next process (SSO), giving priority to the workpiece with the shortest remaining processing time (SRPT), giving priority to the workpiece with the shortest total processing time (SRM), giving priority to the workpiece with the fastest processing completion (SLP), giving priority to the workpiece with the lowest machine tool utilization (SMP), giving priority to the workpiece with the longest remaining processing time (LOP), giving priority to the workpiece with the long processing time in the next process (LSO), giving priority to the workpiece with the long remaining processing time (LRPT), giving priority to the workpiece with the longest total processing time (LRM), giving priority to the workpiece with the slowest processing completion (LLP), and giving priority to the workpiece with the highest machine tool utilization (LMP). The reward function evaluates the equipment utilization change difference caused by a decision after the decision is made, which is expressed as:

[0063] r t =mr t -mr t-1 (6)

[0064] Where r t For the reward of making a decision, mr t is the sum of all equipment utilization after the decision, mr t-1 It is the sum of all equipment utilization before the decision is made.

[0065] Agent refers to the agent in the reinforcement learning algorithm, which is used to select rules. In reinforcement learning, the agent iterates through interaction with the environment state module. The interaction process includes: obtaining the environment state, making choices based on multiple rules, obtaining rewards under the choice, and updating the environment state. The parameters of the agent are optimized through iteration.

[0066] The disturbance identification and dynamic scheduling module is used to identify the disturbance type in the production process, and make production scheduling decisions based on the disturbance situation, and finally output the workpiece JSON data block and the equipment JSON data block. The workpiece JSON data block contains 6 keys, i represents the workpiece index, num represents the total number of workpiece processes, cur represents the process index currently completed by the workpiece, i_start represents the start time of each process of the workpiece, i_end represents the end time of each process of the workpiece, and mac represents the equipment index assigned to each process of the workpiece. The equipment JSON data block contains 4 keys, j represents the equipment index, j_start represents the start time of all processes assigned to the equipment, j_end represents the end time of all processes assigned to the equipment, and job represents the workpiece index to which all processes assigned to the equipment belong.

[0067] The implementation process of the dynamic scheduling method for lightweight information interaction in digital twin workshops is as follows:

[0068] Preparation stage: The production process is simulated on the digital twin workshop platform. The intelligent agent in the high-response dynamic scheduling module interacts with the environmental state module associated with the scheduling. The intelligent agent obtains the environmental state and makes decisions based on 12 rules. The decision changes the environmental state and obtains the reward under the decision. By repeating the above process, the arrangement of all processes is completed, and the intelligent agent parameters are optimized in multiple iterations to obtain the trained dynamic scheduling intelligent agent.

[0069] Step 1: The digital twin workshop platform provides real-time monitoring data for the equipment status module, workpiece status module, and scheduling-related environment status module;

[0070] Step 2: When an abnormal event occurs during the production process, the disturbance identification and dynamic scheduling module immediately obtains the current equipment status module M and the current workpiece status module J n , judge the currently completed process, the process being completed, and the process to be completed, determine the dynamic scheduling time according to the immediate disturbance and the delayed disturbance, call the intelligent agent to make a quick decision on the process to be completed based on the current equipment status, and output the workpiece JSON data block and the equipment JSON data block.

[0071] Step three: The workpiece JSON data block and the equipment JSON data block can be directly used as lightweight data carriers for simulation and decision-making of production scheduling on the digital twin workshop platform to achieve high-response dynamic scheduling of the production process.

Claims

1. A dynamic scheduling method for lightweight information interaction of digital twin workshops, wherein the system implementing the method comprises an equipment status module, a workpiece status module, a scheduling-related environment status module, a high-response dynamic scheduling module, and a disturbance identification and dynamic scheduling module; characterized in that: The intelligent agent in the high-response dynamic scheduling module interacts with the scheduling-related environmental status module to train a dynamic scheduling intelligent agent; based on the digital twin workshop platform, real-time monitoring data is provided for the equipment status module, workpiece status module, and scheduling-related environmental status module; When an abnormal event occurs during the production process, the disturbance identification and dynamic scheduling module immediately obtains the current equipment status module M and the current workpiece status module J n , judge the currently completed process, the process being completed, and the process to be completed, determine the dynamic scheduling time according to the immediate disturbance and the delayed disturbance, call the intelligent agent to make a quick decision on the process to be completed in combination with the current equipment status, and output the workpiece JSON data block and the equipment JSON data block; the workpiece JSON data block and the equipment JSON data block are directly used as lightweight data carriers for the simulation and decision-making of production scheduling on the digital twin workshop platform, so as to realize high-response dynamic scheduling of the production process.

2. According to claim 1, a dynamic scheduling method for lightweight information interaction of digital twin workshops is characterized in that: The equipment status module is used to reflect the status of a single generated equipment. For multiple equipment in the workshop, multiple equipment modules can be generated. The total number of equipment is set to m, m ≥ 1, and the equipment index is set to j, j ∈ [1, m]. Each equipment module contains: equipment index j, the start time set of the processing operation assigned to the equipment The end time set of the processing operations assigned to the equipment Index of the workpiece assigned to the processing operation on the machine 3. According to claim 1, a dynamic scheduling method for lightweight information interaction of digital twin workshops is characterized in that: The workpiece status module is used to reflect the status of a single workpiece. The total number of workpieces is set to n, n ≥ 1, and the workpiece index is set to i, i ∈ [1, n]. Each workpiece module contains: workpiece index i, the total number of workpiece processes The number of processes currently completed by the workpiece The start time of each process of the workpiece The end time of each process of the workpiece The equipment index set assigned to each process of processing the workpiece 4. According to claim 1, a dynamic scheduling method for lightweight information interaction of digital twin workshops is characterized in that: The scheduling-related environment status module is used to obtain the workshop status in real time through the digital twin workshop platform, which includes 10 parameters at the three levels of equipment, workpiece and processing for monitoring the production workshop status, namely, the average equipment utilization rate s mu , the average number of processes arranged for each machine tool s mp , the average probability of workpieces being ju , average process completion rate s jp , the workpiece number s corresponding to the current production task jn , the process number of the current production workpiece jo , the processing machine number of the current workpiece jm , the unprocessed process s of the current workpiece jr , the processing time of the current workpiece s jt 、The processing time of the current equipment s mt .

5. A dynamic scheduling method for lightweight information interaction of digital twin workshops according to claim 1, characterized in that: The high-response dynamic scheduling module includes 12 decision rules, reward functions, and intelligent agents, wherein the decision rules include: giving priority to the workpiece SOP with the least remaining processes, giving priority to the workpiece SSO with a short processing time for the next process, giving priority to the workpiece SRPT with a short remaining processing time, giving priority to the workpiece SRM with the shortest total processing time, giving priority to the workpiece SLP with the fastest processing completion, giving priority to the workpiece SMP with the lowest machine tool utilization, giving priority to the workpiece LOP with the longest remaining processing time, giving priority to the workpiece LSO with a long processing time for the next process, giving priority to the workpiece LRPT with a long remaining processing time, giving priority to the workpiece LRM with the longest total processing time, giving priority to the workpiece LLP with the slowest processing completion, and giving priority to the workpiece LMP with the highest machine tool utilization; the reward function is the difference in equipment utilization; the intelligent agent refers to the intelligent agent in the reinforcement learning algorithm, which is used to select rules.

6. A dynamic scheduling method for lightweight information interaction of digital twin workshops according to claim 1, characterized in that: The disturbance identification and dynamic scheduling module is used to identify the disturbance type in the production process, and make production scheduling decisions according to the disturbance situation, and finally output the workpiece JSON data block and the equipment JSON data block; wherein the workpiece JSON data block contains 6 keys, i represents the workpiece index, num represents the total number of workpiece processes, cur represents the process index currently completed by the workpiece, i_start represents the start time of each process of the workpiece, i_end represents the end time of each process of the workpiece, and mac represents the equipment index assigned to each process of the workpiece; the equipment JSON data block contains 4 keys, j represents the equipment index, j_start represents the start time of all processes assigned to the equipment, j_end represents the end time of all processes assigned to the equipment, and job represents the workpiece index to which all processes assigned to the equipment belong.

Images