Edge computing deployment method oriented to manufacturing shop communication load balance

Abstract

The invention provides an edge computing deployment method for communication load balance of a manufacturing workshop in the technical field of new-generation information technology and high-end manufacturing equipment, the manufacturing workshop is divided into a processing area, a control area and a management area, the available computing power of each piece of edge computing equipment is quantified by counting the total amount of computing tasks in each area, and the computing capacity of each piece of edge computing equipment is calculated. The method comprises the following steps: establishing a distribution relation between tasks and equipment, analyzing transmission time consumption and calculation time consumption of a manufacturing workshop calculation task by utilizing calculation and transmission advantages of edge calculation at a near equipment end, and evaluating total completion time and total energy consumption for completing a total calculation task; and finally, designing a task scheduling algorithm for improving deep reinforcement learning and an ant colony optimization algorithm by taking the calculation task completion time, the energy consumption of the edge calculation equipment and the load balance rate as sum evaluation indexes, solving an optimal scheduling objective function and realizing an optimal deployment scheme of the edge calculation equipment in a manufacturing workshop. And the network stability and reliability of a manufacturing workshop are improved.

Description

Technical Field

[0001] The present invention relates to the field of new generation information technology and high-end manufacturing equipment technology, and specifically to an edge computing deployment method for communication load balancing in a manufacturing workshop. Background Art

[0002] With the rapid development of the manufacturing industry, the communication load inside the manufacturing workshop continues to increase, and the types of tasks to be processed and the amount of real-time sensor data have increased significantly. In order to improve the production efficiency of the workshop and ensure the operating stability of equipment such as CNC machine tools, machining centers, and logistics forklifts, it is urgent to jointly optimize the edge computing technology and load balancing strategy for the manufacturing workshop communication network, and schedule massive data processing tasks and delay-sensitive tasks to the edge computing devices close to the data source, reduce the transmission delay of the overall network and improve the response speed of computing tasks, and effectively solve the challenges of uneven distribution of computing and storage resources in the workshop, limited communication bandwidth interaction between devices, and weak real-time data processing capabilities. Therefore, by developing edge computing deployment technology for manufacturing workshop communication load balancing, not only can the efficient use of distributed computing resources at the manufacturing workshop level be achieved, but also the dual needs of real-time and energy consumption management in industrial tasks will be met, laying a solid foundation for the realization of intelligent manufacturing.

[0003] At present, the edge computing technology for optimizing communication computing networks mainly focuses on theoretical node resource allocation and static load balancing, lacking comprehensive consideration of the actual characteristics of the manufacturing workshop and the load conditions of industrial tasks, which makes it difficult for each edge computing node to fully realize the real-time interactive support and task execution guarantee of the communication computing network of the manufacturing workshop. Aslanpour et al. proposed a serverless edge computing load balancing method based on empirical weight adjustment in the international high-level paper "Load balancing for heterogeneous serverless edge computing: A performance-driven and empirical approach." (Future Generation Computer Systems, Vol. 154, 2024, pp. 266-280). The above method is essentially still aimed at optimizing the edge computing device placement plan for static load balancing, and lacks the ability to process and dynamically optimize real-time data tasks in industrial field automation. Therefore, it is necessary to fully consider the load balancing relationship between the communication computing capacity and the computing tasks to be executed in the edge computing network in the manufacturing workshop, optimize the task matching and resource allocation relationship between computing tasks and edge computing in different areas of the manufacturing workshop, and ensure the execution efficiency of large-scale concurrent computing tasks. Summary of the invention

[0004] The purpose of the present invention is to provide an edge computing deployment method for communication load balancing in a manufacturing workshop. The manufacturing workshop is divided into three areas: processing, control and management. The total amount of computing tasks in each area is counted, the available computing capacity of each edge computing device is quantified, and the distribution relationship between tasks and devices is established. At the same time, the computing and transmission advantages of edge computing near the device are utilized to analyze the transmission time and computing time of computing tasks in the manufacturing workshop, and the total completion time and total energy consumption of the overall computing tasks are evaluated. The computing task completion time, edge computing device energy consumption and load balancing rate are used as the total evaluation indicators. Finally, an improved task scheduling algorithm of deep reinforcement learning and ant colony optimization algorithm is designed to solve the optimal scheduling objective function and realize the optimal deployment plan of edge computing devices in the manufacturing workshop, so as to solve the resource efficiency and network load bottlenecks of the manufacturing workshop when scheduling large-scale complex computing tasks, which has important application value.

[0005] To achieve the above object, the technical solution adopted by the present invention is to provide an edge computing deployment method for manufacturing workshop communication load balancing, which includes the following steps:

[0006] Step S1: Focusing on the equipment inside the manufacturing workshop, divide the manufacturing workshop into processing areas according to business characteristics and operation processes. , control area , Management area , and then count the total amount of computing tasks in all regions and record it as ;

[0007] Step S2: Define the edge computing devices available in the manufacturing workshop as ;

[0008] Step S3: Define the deployment allocation relationship between computing tasks in the manufacturing workshop and edge computing devices as ,in, , ,like , represents the computing tasks in the manufacturing workshop Assigned to edge computing devices To execute, if , then it represents Not on edge computing devices Execute on;

[0009] Step S4: Define the data transmission time of computing tasks in different areas of the manufacturing workshop Compute time with tasks Different manufacturing workshop areas will have different task data volumes and communication link capabilities. The data transmission time is determined by the task data volume and the communication link transmission rate, which can be quantitatively expressed as: ,in, To complete the computational task The number of instructions required, To calculate the transmission rate of the task uplink and given by Shannon's theorem, , is the uplink bandwidth, For computing tasks The channel gain between the manufacturing area and the edge computing device, Transmit power for the link in the manufacturing workshop, is the noise power existing in the link bandwidth of the manufacturing workshop; the task calculation time is quantitatively expressed as ,in For edge computing devices The computing power of For edge computing devices The proportion of currently available computing resources;

[0010] Step S5: Calculate the computing tasks in the manufacturing workshop Total completion time , computing is arranged on edge computing devices The total completion time of all manufacturing workshops on the ;

[0011] Step S6: Define the task transmission energy consumption of computing tasks in different areas of the manufacturing workshop Calculate energy consumption with tasks ,in, , For edge computing devices The tail energy consumed after transmission; Calculate the energy consumption for the task, For edge computing devices The energy consumption generated by each CPU cycle;

[0012] Step S7: Calculate the computing tasks in the manufacturing workshop Total energy consumption , computing is arranged on edge computing devices The total energy consumption of all manufacturing workshop computing tasks on ;

[0013] Step S8: Define the allocation to the edge computing server The number of tasks on , get the average number of tasks assigned to all edge computing servers At the same time, the average computing time of all edge computing servers executing computing tasks in all manufacturing workshops is obtained as , design the load balancing rate index of the edge computing server when executing the computing tasks of the entire manufacturing workshop ,in The weight of task allocation balance indicates the contribution of task allocation balance to the overall load balance. is the weight of duration balance, indicating the contribution of calculation duration balance to the overall load balance. and satisfy ;

[0014] Step S9: Design the edge computing deployment scheduling optimization objective function that comprehensively considers task computing time, task computing energy consumption and load balancing rate: ,in, is the weight of the load balancing ratio, is the weight of the task calculation duration, is the weight of the task computing energy consumption, satisfying ;

[0015] Step S10: Design an optimization solution algorithm driven by the ant colony algorithm and the deep reinforcement learning network to solve the edge computing deployment scheduling solution that comprehensively considers the task computing time, task computing energy consumption and load balancing rate, and outputs the initial optimization deployment solution through the deep reinforcement learning network module, where the state set is , the action set is , the reward is to optimize the value of the objective function, and the network training is completed through experience replay and delayed soft update mechanism;

[0016] Step S11: Set the pheromone factor, heuristic factor, and pheromone volatilization factor in the genetic algorithm, and solve the deep reinforcement learning As the initial pheromone concentration of the ant colony algorithm, Design is the inspiration factor, construction and The inversely proportional function is used as the heuristic function of the ant colony algorithm, and the optimal edge computing device deployment solution is solved by a given number of iterations;

[0017] Step S12: Apply the edge computing device deployment solution obtained to the manufacturing workshop site, and perform edge computing on the processing areas , control area , Management area Perform computing task scheduling with communication load balancing.

[0018] Furthermore, the edge computing device in step S2 Number of and the number of computational tasks The relationship between , so that each computing task is assigned to at least one edge computing device.

[0019] Furthermore, the computing tasks in different areas of the manufacturing workshop in step S4 have different task data volumes and communication link capabilities.

[0020] Furthermore, the data transmission in step S4 takes time Compute time with tasks The overall time It is used to evaluate the overall execution efficiency of computing tasks and provide a basis for task scheduling and resource allocation.

[0021] Compared with the prior art, the present invention has the following beneficial effects: the invented edge computing deployment method for communication load balancing in manufacturing workshops combines the computing and transmission advantages of edge computing near the device end, takes the computing task completion time, edge computing device energy consumption and load balancing rate as the total evaluation index, and designs an improved task scheduling algorithm of deep reinforcement learning and ant colony optimization algorithm, which effectively reduces the execution time of complex computing tasks in the manufacturing workshop, reduces the unit power consumption of edge computing devices, and improves the network stability and reliability of the manufacturing workshop. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flow chart for implementing the edge computing deployment method for manufacturing workshop communication load balancing according to the present invention.

[0023] Figure 2 This is a schematic diagram of the edge computing deployment optimization model that takes into account task computing duration, task computing energy consumption and load balancing rate in the present invention. DETAILED DESCRIPTION

[0024] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0025] See also Figure 1 and Figure 2 The present invention provides an edge computing deployment method for manufacturing workshop communication load balancing, comprising the following steps:

[0026] Step S1: Focusing on the equipment inside the manufacturing workshop, divide the manufacturing workshop into processing areas according to business characteristics and operation processes. , control area , Management area , and then count the total amount of computing tasks in all regions and record it as ;

[0027] Step S2: Define the edge computing devices available in the manufacturing workshop as ;

[0028] Step S3: Define the deployment allocation relationship between computing tasks in the manufacturing workshop and edge computing devices as ,in, , ,like , represents the computing tasks in the manufacturing workshop Assigned to edge computing devices To execute, if , then it represents Not on edge computing devices Execute on;

[0029] Step S4: Define the data transmission time of computing tasks in different areas of the manufacturing workshop Compute time with tasks Different manufacturing workshop areas will have different task data volumes and communication link capabilities. The data transmission time is determined by the task data volume and the communication link transmission rate, which can be quantitatively expressed as: ,in, To complete the computational task The number of instructions required, , is the uplink bandwidth, For computing tasks The channel gain between the manufacturing area and the edge computing device, Transmit power for the link in the manufacturing workshop, is the noise power existing in the link bandwidth of the manufacturing workshop; the task calculation time is quantitatively expressed as ,in For edge computing devices The computing power of For edge computing devices The proportion of currently available computing resources;

[0030] Step S5: Calculate the computing tasks in the manufacturing workshop Total completion time , computing is arranged on edge computing devices The total completion time of all manufacturing workshops on the ;

[0031] Step S6: Define the task transmission energy consumption of computing tasks in different areas of the manufacturing workshop Calculate energy consumption with tasks ,in, , For edge computing devices The tail energy consumed after transmission; Calculate the energy consumption for the task, For edge computing devices The energy consumption generated by each CPU cycle;

[0032] Step S7: Calculate the computing tasks in the manufacturing workshop Total energy consumption , computing is arranged on edge computing devices The total energy consumption of all manufacturing workshop computing tasks on ;

[0033] Step S8: Define the allocation to the edge computing server The number of tasks on , get the average number of tasks assigned to all edge computing servers At the same time, the average computing time of all edge computing servers executing computing tasks in all manufacturing workshops is obtained as , design the load balancing rate index of the edge computing server when executing the computing tasks of the entire manufacturing workshop ,in The weight of task allocation balance indicates the contribution of task allocation balance to the overall load balance. is the weight of duration balance, indicating the contribution of calculation duration balance to the overall load balance. and satisfy ;

[0034] Step S9: Design the edge computing deployment scheduling optimization objective function that comprehensively considers task computing time, task computing energy consumption and load balancing rate: ,in, is the weight of the load balancing ratio, is the weight of the task calculation duration, is the weight of the task computing energy consumption, satisfying ;

[0035] Step S10: Design an optimization solution algorithm driven by the ant colony algorithm and the deep reinforcement learning network to solve the edge computing deployment scheduling scheme that comprehensively considers the task computing time, task computing energy consumption and load balancing rate, and outputs the initial optimization deployment scheme through the deep reinforcement learning network module, where the state set is , the action set is , the reward is to optimize the value of the objective function, and the network training is completed through experience replay and delayed soft update mechanism;

[0036] Step S11: Set the pheromone factor, heuristic factor, and pheromone volatilization factor in the genetic algorithm, and solve the deep reinforcement learning As the initial pheromone concentration of the ant colony algorithm, Design is the inspiration factor, construction and The inversely proportional function is used as the heuristic function of the ant colony algorithm, and the optimal edge computing device deployment solution is solved by a given number of iterations;

[0037] Step S12: Apply the edge computing device deployment solution obtained to the manufacturing workshop site, and perform edge computing on the processing areas , control area , Management area Perform computing task scheduling with communication load balancing.

[0038] The following takes a dust-free, constant temperature CNC machine tool manufacturing workshop containing 5 edge computing devices as an example to further illustrate the specific implementation and effects of the invention.

[0039] The manufacturing workshop needs to calculate and process 30 complex computing tasks, including production order scheduling, remote processing instruction execution, parts inventory inspection, and semi-finished product quality control. The number of computing task instructions required is between 3000 and 5000 MI (Million Instructions), and the computing task data size is between 500 and 100 MB.

[0040] Table 1 provides a comparison of the total time taken by the method of the present invention to complete all computing tasks in each area of ​​the manufacturing workshop (unit: seconds). The comparison results show that the method of the present invention fully evaluates the processing capability of the edge computing device and optimizes the dynamic scheduling allocation of computing tasks in the manufacturing workshop of the edge computing device.

[0041] Table 1 Comparison of the total time taken by the method of the present invention and other methods to complete all computing tasks in each area of ​​the manufacturing workshop

[0042]

[0043] Table 2 provides the total energy consumption (unit: watt) when the method of the present invention is deployed for edge computing load balancing in manufacturing workshops. The comparison results show that the method of the present invention can fully call on the idle resources of edge computing devices, reduce additional data transmission losses, and reduce the overall network communication and on-site computing energy consumption losses in the manufacturing workshop.

[0044] Table 2 Overall energy consumption of the edge computing deployment for manufacturing workshop communication load balancing according to the method of the present invention (unit: watt)

[0045]

[0046] The present invention provides an edge computing deployment method for communication load balancing in manufacturing workshops, which defines the association and computing power of available edge computing devices in the manufacturing workshop, quantifies the transmission time and computing time of computing tasks, evaluates the total completion time and total energy consumption of computing tasks in the manufacturing workshop, designs a load balancing rate indicator, and realizes optimized scheduling based on ant colony algorithm and deep reinforcement learning. The optimal deployment scheme of edge computing devices solved has significant advantages in improving task scheduling efficiency and load balancing capability.

[0047] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art may make some modifications and improvements without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be based on the definition of the claims.

Claims

1. An edge computing deployment method for manufacturing workshop communication load balancing, characterized in that: The following steps are involved: Step S1: Focusing on the equipment inside the manufacturing workshop, divide the manufacturing workshop into processing areas according to business characteristics and operation processes. , control area , Management area , and then count the total amount of computing tasks in all regions and record it as ; Step S2: Define the edge computing devices available in the manufacturing workshop as ; Step S3: Define the deployment allocation relationship between computing tasks in the manufacturing workshop and edge computing devices as ,in, , ,like , represents the computing tasks in the manufacturing workshop Assigned to edge computing devices To execute, if , then it represents Not on edge computing devices Execute on; Step S4: Define the data transmission time of computing tasks in different areas of the manufacturing workshop Compute time with tasks Different manufacturing workshop areas will have different task data volumes and communication link capabilities. The data transmission time is determined by the task data volume and the communication link transmission rate, which can be quantitatively expressed as: ,in, To complete the computational task The number of instructions required, , is the uplink bandwidth, For computing tasks The channel gain between the manufacturing area and the edge computing device, Transmit power for the link in the manufacturing workshop, is the noise power existing in the link bandwidth of the manufacturing workshop; the task calculation time is quantitatively expressed as ,in For edge computing devices The computing power of For edge computing devices The proportion of currently available computing resources; Step S5: Calculate the computing tasks in the manufacturing workshop Total completion time , computing is arranged on edge computing devices The total completion time of all manufacturing workshops on the ; Step S6: Define the task transmission energy consumption of computing tasks in different areas of the manufacturing workshop Calculate energy consumption with tasks ,in, , For edge computing devices The tail energy consumed after transmission; Calculate the energy consumption for the task, For edge computing devices The energy consumption generated by each CPU cycle; Step S7: Calculate the computing tasks in the manufacturing workshop Total energy consumption , computing is arranged on edge computing devices The total energy consumption of all manufacturing workshop computing tasks on ; Step S8: Define the allocation to the edge computing server The number of tasks on , get the average number of tasks assigned to all edge computing servers At the same time, the average computing time of all edge computing servers executing computing tasks in all manufacturing workshops is obtained as , design the load balancing rate index of the edge computing server when executing the computing tasks of the entire manufacturing workshop ,in The weight of task allocation balance indicates the contribution of task allocation balance to the overall load balance. is the weight of duration balance, indicating the contribution of calculation duration balance to the overall load balance. and satisfy ; Step S9: Design the edge computing deployment scheduling optimization objective function that comprehensively considers task computing time, task computing energy consumption and load balancing rate: ,in, is the weight of the load balancing ratio, is the weight of the task calculation duration, is the weight of the task computing energy consumption, satisfying ; Step S10: Design an optimization solution algorithm driven by the ant colony algorithm and the deep reinforcement learning network to solve the edge computing deployment scheduling scheme that comprehensively considers the task computing time, task computing energy consumption and load balancing rate, and outputs the initial optimization deployment scheme through the deep reinforcement learning network module, where the state set is , the action set is , the reward is to optimize the value of the objective function, and the network training is completed through experience replay and delayed soft update mechanism; Step S11: Set the pheromone factor, heuristic factor, and pheromone volatilization factor in the genetic algorithm, and solve the deep reinforcement learning As the initial pheromone concentration of the ant colony algorithm, Design is the inspiration factor, construction and The inversely proportional function is used as the heuristic function of the ant colony algorithm, and the optimal edge computing device deployment solution is solved by a given number of iterations; Step S12: Apply the edge computing device deployment solution obtained to the manufacturing workshop site, and perform edge computing on the processing areas , control area , Management area Perform computing task scheduling with communication load balancing.

2. The edge computing deployment method according to claim 1, characterized in that: The edge computing device in step S2 Number of and the number of computational tasks The relationship between , so that each computing task is assigned to at least one edge computing device.

3. The edge computing deployment method according to claim 1, characterized in that: The computing tasks in different areas of the manufacturing workshop in step S4 have different task data volumes and communication link capabilities. The number of computing task instruction requirements is between 3000 and 5000 MI, and the computing task data size is between 500 and 100 MB.

4. The edge computing deployment method according to claim 1, characterized in that: The data transmission in step S4 takes time Compute time with tasks The overall time It is used to evaluate the overall execution efficiency of computing tasks and provide a basis for task scheduling and resource allocation.

Images