Numerical control machine tool calculation task dynamic scheduling method based on edge calculation

Abstract

The invention discloses a numerical control machine tool calculation task dynamic scheduling method based on edge calculation, which comprises the following steps: defining an equipment set containing a plurality of numerical control machine tools and a discretized industrial field time sequence, acquiring a to-be-executed calculation task of each machine tool in real time, and splitting the to-be-executed calculation task into sub-calculation task sequences with a front-back order relation; quantifying the processing duration of the task in the numerical control system and the edge computing facility, and setting an optimization scheduling target; a Markov decision process and a deep reinforcement learning network are combined, and a double-delay depth deterministic strategy is adopted to realize dynamic scheduling. According to the numerical control machine tool calculation task dynamic scheduling method based on edge calculation, the resource efficiency and the response speed of the numerical control machine tool during complex calculation task processing are improved, and an effective guarantee is provided for high-precision machining of the numerical control machine tool.

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 a method for dynamic scheduling of computing tasks for CNC machine tools based on edge computing. Background Art

[0002] CNC machine tools, as pillars of the high-end equipment manufacturing industry, have a machining accuracy that directly determines the overall performance of complex components such as aircraft landing gear, turbine blades, and engine blocks. Furthermore, with the rapid development of new-generation information technologies such as 5G networks, the Internet of Things, big data, and artificial intelligence, new CNC applications designed to enhance CNC machine tool machining accuracy are constantly emerging. The computational complexity of tasks that CNC machine tools and their underlying CNC systems must handle in real time is increasing exponentially. This leads to severe computing challenges for CNC machine tools in industrial sites when handling computationally intensive and latency-sensitive applications, which in turn impacts the controllability of product machining accuracy and the stability of their operational performance. To compensate for the lack of local computing power and the high cost of cloud computing resources, it is imperative to further explore the computing and communication advantages of edge computing infrastructure near the machine tool. Dynamic scheduling optimization techniques tailored to the characteristics of CNC system computational tasks and CNC machine tool computational resources are being developed to reduce the latency consumed by sensor data transmission and simultaneously improve the computational efficiency of real-time response in CNC programs.

[0003] While current edge computing-enhanced CNC machine tools and systems achieve significant static configuration enhancements in computing resources and bandwidth capabilities, they lack the ability to dynamically optimize scheduling based on the associated characteristics and logical relationships of specific computing tasks. This hinders the full potential of edge computing facilities in enhancing the interaction between CNC machine tools and systems. In their internationally recognized paper, "Edge intelligence-driven digital twin of CNC system: Architecture and deployment" (Robotics and Computer-Integrated Manufacturing, Vol. 79, 2023, pp. 102-418), Yu et al. proposed a theoretical modeling approach based on the hierarchical structure of CNC systems. In their internationally recognized paper, "STEP-NC enabled edge–cloud collaborative manufacturing system for compliant CNC machining" (Journal of Manufacturing Systems, Vol. 72, 2024, pp. 460-474), Xiao et al. designed a manufacturing system that optimizes CNC systems through collaborative interaction between edge computing and cloud computing, achieving accurate and intelligent sensor data exchange. The above methods are essentially the design of hardware deployment frameworks for collaborative interaction between edge computing and CNC machine tools, and lack the evaluation and utilization of the characteristics of the computing tasks to be executed by CNC machine tools. At the same time, due to the random arrival and high real-time characteristics of complex tasks such as feeding, compensation, and following that need to be executed by industrial field CNC systems, it is difficult to meet the needs of actual process scenarios by relying solely on hardware architecture for unified deployment and processing. Therefore, it is necessary to fully consider the correlation between the computing tasks to be executed by the CNC system, combine the software and hardware resource capabilities between edge computing facilities and CNC machine tools, and invent a real-time scheduling method that can dynamically optimize and adjust computing tasks, optimize the bandwidth allocation and computing resource allocation ratio between CNC machine tools and edge computing, and ensure the quality and efficiency of high-precision machining tasks. Summary of the Invention

[0004] The purpose of the present invention is to provide a dynamic scheduling method for computing tasks of CNC machine tools based on edge computing, define the equipment set of multiple CNC machine tools and the discretized industrial field time series, split the computing tasks into sub-computing task sequences and model the sequence relationship of the computing task sequences, and realize the dynamic response scheduling of computing resources by constructing an optimized scheduling target that quantifies the processing time of computing tasks in CNC systems and edge computing facilities, integrating Markov decision processes with deep reinforcement learning, and deploying a dual-delay deep deterministic strategy.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is to provide a method for dynamic scheduling of computing tasks of CNC machine tools based on edge computing, which includes the following steps:

[0006] S1: Define a machine tool equipment set containing multiple CNC machine tools , set the discretized industrial site time series , where the time interval between each discrete time point is ;

[0007] S2: Based on historical monitoring data and actual order information, obtain the dynamically generated computing tasks to be executed for each CNC machine tool in real time, and define the tasks for each machine tool. , and split the computing task into a sub-computing task sequence according to the process complexity and real-time requirement, and use the triple set to identify each sub-computing task. The data size, required computing resources, and deployment strategy are uniformly represented;

[0008] S3: Define the order relationship between sub-computing task sequences, and analyze the order of the sub-computing task sequences through the constraints of the order of the sub-computing task sequences. The sub-computation task and The relationship between sub-computing tasks;

[0009] S4: For sub-computing tasks processed locally in the CNC system , quantify the total time required to complete the processing of the sub-computation task, check the number of remaining tasks in the CNC system processing queue, and record it through the CNC system timer when the number of remaining tasks is 0 and the sub-task close relationship constraint is met The total time from entering the CNC system queue to completing the CNC system local processing;

[0010] S5: Sub-computing tasks for dynamic scheduling and processing from local CNC systems to edge computing facilities , quantify the time required for wireless data migration and transmission, check the number of remaining tasks in the wireless transmission queue, and record the number of tasks through the CNC system timer when the number of remaining tasks is 0 and the subtask close relationship constraint is met. The total time from entering the wireless transmission queue to completing the wireless transmission process;

[0011] S6: Sub-computing tasks for dynamic scheduling from local CNC systems to edge computing facilities , quantify the time it takes for the sub-computing task to be processed in the edge computing facility, check the number of remaining tasks in the edge computing processing queue, and record it through the edge computing facility timer when the number of remaining tasks is 0 and the sub-task close relationship constraint is met. The total time from entering the edge computing queue to completing edge computing processing;

[0012] S7: Set the goal of optimizing the dynamic scheduling of CNC machine tool computing tasks for edge computing facilities, and achieve this goal by maximizing the number of successful sub-computing tasks completed within the total time;

[0013] S8: Integrate Markov decision processes with deep reinforcement learning networks to characterize the dynamic scheduling process of CNC machine tool computing tasks for edge computing facilities. Design a dual-delay deep deterministic strategy to implement a solution for the scheduling strategy. Improve the state input of the deep reinforcement learning network to form a state set including the number of queued tasks in the CNC system queue, the number of remaining tasks in the CNC system processing queue, the amount of sub-computing task data, and the amount of computing resource required. Improve the state input of the dual-delay deep deterministic strategy to form a state set including the number of queued tasks in the wireless transmission queue, the number of remaining tasks in the wireless transmission queue, the number of queued tasks in the edge computing queue, and the number of remaining tasks in the edge computing processing queue.

[0014] S9: Improve the action input of deep reinforcement learning network to form representation The action set of the overall migration action; improve the action input of the dual-delay deep deterministic strategy to form a bandwidth ratio allocation set , computing resource ratio allocation set A collection of actions;

[0015] S10: Improve the reward value of the deep reinforcement learning network and the double-delay deep deterministic strategy, both defined as the number of successful sub-computation tasks completed in the current time;

[0016] S11: Randomly initialize the model parameters of the deep reinforcement learning network and the double-delay deep deterministic policy, design an update strategy for the Q-value function of the deep reinforcement learning network, use the experience replay mechanism to store each state transition (state, action, reward, next state) of the deep reinforcement learning network and the double-delay deep deterministic policy in the experience pool, and randomly extract small batches of experience samples from the experience pool for policy training;

[0017] S12: After completing a specified number of rounds of sample training for the deep reinforcement learning network and the dual-delay deep deterministic strategy, the network strategy is deployed to the CNC system and edge computing facilities of the CNC machine tool to assist the CNC system in real time in completing the dynamic scheduling and optimal allocation of computing tasks.

[0018] Furthermore, in step S1, the time interval Greater than or equal to 10ms.

[0019] Furthermore, in step S2, the historical monitoring data includes the operating status of the machine tool equipment, fault records and previous task execution status, and the basis for splitting the sub-computing task sequence includes the priority, complexity and required resources of the task.

[0020] Furthermore, in step S3, the sequence analysis of the sub-computing task sequence is based on the task execution time, resource requirements and process requirements to ensure the reasonable scheduling of the sub-computing tasks.

[0021] Furthermore, in step S4, the CNC system timer is used to accurately record the processing duration of the sub-computing task, and the timer resolution is not higher than 10ms.

[0022] Furthermore, in step S5, the numerical control system timer is used to accurately record the wireless transmission duration of the sub-computing task, and the timer resolution is not higher than 10ms.

[0023] Furthermore, in step S6, the edge computing facility timer is used to accurately record the processing time of the sub-computing task, and the timer resolution is not higher than 10ms.

[0024] Furthermore, in step S7, the dynamic scheduling target takes into account the priority, resource requirements and execution time of the sub-computing tasks.

[0025] Furthermore, in step S9 and step S10, the dual-delay deep deterministic strategy dynamically adjusts the scheduling strategy to adapt to different computing task requirements by updating the state input in real time.

[0026] Furthermore, in step S11, the extraction strategy of the small batch of experience samples adopts uniform random sampling or priority experience sampling to enhance the learning effect of the scheduling strategy.

[0027] Compared with the existing technology, the present invention has the following beneficial effects: the present invention proposes a method for dynamic scheduling of computing tasks for CNC machine tools based on edge computing. First, it defines the equipment set of multiple CNC machine tools and the discretized industrial site time series, ensuring real-time acquisition of computing tasks to be executed; secondly, it splits the computing tasks into sub-computing task sequences, and models and characterizes their sequence relationship; then, it constructs an optimized scheduling target for the processing time of quantitative computing tasks in CNC systems and edge computing facilities; finally, it integrates Markov decision processes and deep reinforcement learning, deploys a dual-delay deep deterministic strategy to achieve dynamic scheduling response scheduling. By optimizing task scheduling and resource allocation, this method significantly reduces the resource occupancy of CNC machine tools when processing complex computing tasks, and effectively improves the success rate and resource utilization efficiency of complex CNC computing task scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a flow chart of the method for dynamic scheduling of computing tasks for CNC machine tools based on edge computing of the present invention.

[0029] Figure 2 Schematic diagram of the optimized dynamic scheduling target for CNC machine tool computing tasks for edge computing facilities of the present invention.

[0030] Figure 3 The results of dynamic scheduling of computing tasks for CNC machine tools using the method proposed in this invention and other comparative methods are shown. DETAILED DESCRIPTION

[0031] The present invention will be further described below with reference to the accompanying drawings and examples.

[0032] To address the resource efficiency bottleneck of CNC machine tools when processing complex computing tasks, and to significantly reduce the processing time of CNC system tasks and ensure the response speed of CNC machine tool processing requirements by collaboratively optimizing dynamic task scheduling and resource adjustment allocation between CNC systems and edge computing facilities, this paper proposes a dynamic scheduling method for CNC machine tool computing tasks based on edge computing. By defining a device set including multiple CNC machine tools and a discrete industrial site time series, the method obtains the pending computing tasks of each machine tool in real time and splits them into a sequence of sub-computing tasks with a pre- and post-sequential relationship. The processing time of the tasks in the CNC system and edge computing facilities is quantified, and the optimization scheduling goal is set. The Markov decision process and deep reinforcement learning network are integrated to form a dynamic scheduling mechanism with a double-delay deep deterministic strategy, which effectively improves the success rate and resource utilization efficiency of the scheduling of complex computing tasks for CNC machine tools. Figure 1 The present invention provides a method for dynamically scheduling computing tasks of CNC machine tools based on edge computing, comprising the following steps:

[0033] Step 1: Define a machine tool equipment set containing multiple CNC machine tools , is the total number of machine tools in the manufacturing system; defines the discretized industrial site time series , where the interval between each discrete time point is .

[0034] Step 2: Based on historical monitoring data and actual order information, obtain the dynamically generated pending computing tasks for each CNC machine tool in real time, and define the machine tool equipment ,in It can be divided into sub-computing task sequences according to the complexity of the process and the degree of real-time requirements. A triple set Fully express, in which Indicates completion The amount of data that needs to be processed, For processing The computing resources required, is a set of two tuples, if It means is dispatched to the edge computing facility for execution. , it means that it is executed locally on the CNC system of the CNC machine tool.

[0035] Step 3: Define the order relationship between sub-computing task sequences and pass Indicates the The sub-computation task and The relationship between the sub-computing tasks ( ,like Need to After completion, it can be executed, then define ,like Need to After completion, it can be executed, then define ,like and There is no order relationship between them, so define .

[0036] Step 4: Sub-computing tasks to be processed locally on the CNC system , quantify the total time required for the sub-computation task to complete the processing, and check the number of remaining tasks in the CNC system processing queue If the number of remaining tasks is 0, search the CNC system queue, wireless transmission queue, and edge computing queue to see if there is a task that matches the number of remaining tasks. There exists a Established sub-computing tasks , if there exists such that Established sub-computing tasks , then Still placed in the CNC system queue; on the contrary, if there is no Established sub-computing tasks , then Place it in the CNC system processing queue and define the processing time as , Indicates machine tool equipment The computing power of the CNC system is recorded by the CNC system timer The initial point when the queue is placed in the CNC system processing queue and waiting time .

[0037] Step 5: Dynamically schedule sub-computing tasks that are migrated from the CNC system to the edge computing facility , quantify the time required for wireless data migration and transmission, and check the number of remaining tasks in the wireless transmission queue If the number of remaining tasks is 0, search the CNC system queue, wireless transmission queue, and edge computing queue to see if there is a task that matches the number of remaining tasks. There exists a Established sub-computing tasks , if there exists such that Established sub-computing tasks , then Still placed in the wireless transmission queue; on the contrary, if there is no Established sub-computing tasks , then Place it in the wireless transmission processing queue and define the transmission time as ,in is the data transmission rate, which can be expressed as , is the total channel bandwidth resource, For machine tools The transmission power, is Gaussian power white noise, is the channel gain factor between the machine tool CNC system and the edge computing facility, is the loss factor during wireless transmission, for Time allocated to machine tools The bandwidth ratio is recorded by the CNC system timer The initial point in time when the wireless transmission queue is placed in the wireless transmission processing queue and waiting time .

[0038] Step 6: Dynamically schedule sub-computing tasks that are migrated from the CNC system to the edge computing facility , quantify the time it takes for the sub-computing task to be processed in the edge computing facility, and check the number of remaining tasks in the edge computing processing queue If the number of remaining tasks is 0, search the CNC system queue, wireless transmission queue, and edge computing queue to see if there is a task that matches the number of remaining tasks. There exists a Established sub-computing tasks , if there exists such that Established sub-computing tasks , then Still placed in the edge computing queue; on the contrary, if there is no Established sub-computing tasks , then Place it in the edge computing processing queue and define the edge computing processing time as , and recorded by the edge computing facility timer The waiting time from the edge computing queue to the edge computing processing queue .

[0039] Step 7: Further reference Figure 2 , Figure 2 Provides an optimized dynamic scheduling target for CNC machine tool computing tasks for edge computing facilities. The scheduling target is to maximize the number of successful sub-computing tasks completed within the total time. , we further introduce the following six optimization constraints that need to be satisfied by the maximization objective:

[0040] 1. Constraint 1 is , representing a sub-computing task The specific point in time at which processing is completed within the CNC system;

[0041] 2. Constraint 2 is , representing a sub-computing task The specific point in time when processing is completed in the edge computing facility;

[0042] 3. Constraint 3 is , represents any sub-computation task The successful completion time must be less than the total time ,in ;

[0043] 4. Constraint 4 is , indicating that the bandwidth ratio allocated to all CNC machine tools during wireless transmission of computing task data is less than or equal to 1;

[0044] 5. Constraint 5 is , indicating that the sum of the proportions of computing resources allocated to all CNC machine tools at the edge computing facility is less than or equal to 1;

[0045] 6. Constraint 6 is , indicating that the computing resources allocated at the edge computing facility need to be greater than The computing resources required.

[0046] Step 8: By integrating the Markov decision process with the deep reinforcement learning network, we characterize the dynamic scheduling process of CNC machine tool computing tasks for edge computing facilities, design a double-delay deep deterministic strategy to solve the scheduling strategy, and improve the state input of the deep reinforcement learning network, including the number of queued tasks in the CNC system queue. , the number of remaining tasks in the CNC system processing queue 、 The amount of task data 、 Computing resource requirements ; Define the state input of the double-delay deep deterministic strategy, including the number of queued tasks in the wireless transmission queue , the number of remaining tasks in the wireless transmission queue , the number of queued tasks in the edge computing queue , the number of remaining tasks in the edge computing processing queue .

[0047] Step 9: Improve the action input of the deep reinforcement learning network Whether to migrate to edge computing facilities is represented by binary, thus forming an action set express The overall migration action of ; the action input of the double-delay deep deterministic strategy is defined as the bandwidth ratio allocation set , computing resource ratio allocation set .

[0048] Step 10: Improve the reward values of the deep reinforcement learning network and the double-delay deep deterministic strategy, both defined as the current time The number of successful sub-computation tasks completed within .

[0049] Step 11: Randomly initialize the model parameters of the deep reinforcement learning network and the double-delay deep deterministic strategy, design the update strategy of the deep reinforcement learning network Q value function, use the experience replay mechanism to store each state transition (state, action, reward, next state) of the deep reinforcement learning network and the double-delay deep deterministic strategy in the experience pool, and randomly extract small batches of experience samples from the experience pool for strategy training. In single-time training, first input 、 、 、 To double-delay deep deterministic strategy, get 、 The output value of further combines the above state parameters with 、 、 The latest state set is formed and the reward value is obtained interactively. Combined with the next state set, it is uniformly input into the deep reinforcement learning network to update the Q value. Backpropagation and delayed soft update are performed simultaneously to iteratively optimize the training parameters in the deep reinforcement learning network and the double-delay deep deterministic strategy.

[0050] Step 12: After completing the specified number of rounds of sample training for the deep reinforcement learning network and the dual-delay deep deterministic strategy, the network strategy is deployed to the CNC system and edge computing facilities of the CNC machine tool to assist the CNC system in dynamic scheduling and optimal allocation of computing tasks in real time.

[0051] The following takes 10 CNC machine tools equipped with Siemens 840D CNC systems and Inspur EIS200 edge computing facilities as an example to further illustrate the specific implementation and effects of the invention.

[0052] For 1000 cases of special-shaped rotary parts processing and tool compensation calculation tasks that randomly and continuously arrive at different task arrival rates in the CNC systems of 10 CNC machine tools within 100 seconds, a comprehensive experimental comparison of the number of successful tasks was conducted between the CNC machine tool calculation task dynamic scheduling method based on edge computing designed by this invention and the use of only deep reinforcement learning networks and the task random allocation strategy and task overall migration strategy currently commonly used in industrial fields. Figure 3 , Figure 3 The effects of the method proposed in the present invention and other comparative methods on dynamic scheduling of CNC machine tool computing tasks are provided, and the indicators of this embodiment are better than all the comparative methods.

[0053] Table 1 provides a comparison of the time-consuming indicators generated by the proposed method for real-time arrival computing tasks of CNC systems, combined with edge computing facilities for real-time strategy reasoning. Compared with scheduling schemes based on heuristic algorithms and static optimization scheduling, this method is fully applicable to the on-site computing resource allocation of CNC machine tools in dynamic environments and meets the real-time requirements of high-precision and high-complexity machining tasks of CNC systems.

[0054] Table 1 Comparison of the time consumption of strategy reasoning for computing tasks of numerical control systems by the method proposed in this invention and other methods.

[0055]

[0056] The present invention provides a dynamic scheduling method for CNC machine tool computing tasks based on edge computing, which combines Markov decision process and deep reinforcement learning, and adopts a dual-delay deep deterministic strategy to achieve dynamic scheduling response. By optimizing task scheduling and resource allocation, it significantly reduces the resource occupancy rate of CNC machine tools when processing complex computing tasks, and effectively improves the success rate and resource utilization efficiency of complex CNC computing task scheduling.

[0057] Although the present invention has been disclosed above with reference to 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. A dynamic scheduling method for computing tasks of CNC machine tools based on edge computing, characterized in that , including the following steps: S1: Define a machine tool equipment set containing multiple CNC machine tools , set the discretized industrial site time series , where the time interval between each discrete time point is ; S2: Based on historical monitoring data and actual order information, obtain the dynamically generated computing tasks to be executed for each CNC machine tool in real time, and define the tasks for each machine tool. , and split the computing task into a sub-computing task sequence according to the process complexity and real-time requirement, and use the triple set to identify each sub-computing task. The data size, required computing resources, and deployment strategy are uniformly represented; S3: Define the order relationship between sub-computing task sequences, and analyze the order of the sub-computing task sequences through the constraints of the order of the sub-computing task sequences. The sub-computation task and The relationship between sub-computing tasks; S4: For sub-computing tasks processed locally in the CNC system , quantify the total time required to complete the processing of the sub-computation task, check the number of remaining tasks in the CNC system processing queue, and record it through the CNC system timer when the number of remaining tasks is 0 and the sub-task close relationship constraint is met The total time from entering the CNC system queue to completing the CNC system local processing; S5: Sub-computing tasks for dynamic scheduling and processing from local CNC systems to edge computing facilities , quantify the time required for wireless data migration and transmission, check the number of remaining tasks in the wireless transmission queue, and record the number of tasks through the CNC system timer when the number of remaining tasks is 0 and the subtask close relationship constraint is met. The total time from entering the wireless transmission queue to completing the wireless transmission process; S6: Sub-computing tasks for dynamic scheduling from local CNC systems to edge computing facilities , quantify the time it takes for the sub-computing task to be processed in the edge computing facility, check the number of remaining tasks in the edge computing processing queue, and record it through the edge computing facility timer when the number of remaining tasks is 0 and the sub-task close relationship constraint is met. The total time from entering the edge computing queue to completing edge computing processing; S7: Set the goal of optimizing the dynamic scheduling of CNC machine tool computing tasks for edge computing facilities, and achieve this goal by maximizing the number of successful sub-computing tasks completed within the total time; S8: Integrate Markov decision processes with deep reinforcement learning networks to characterize the dynamic scheduling process of CNC machine tool computing tasks for edge computing facilities. Design a dual-delay deep deterministic strategy to implement a solution for the scheduling strategy. Improve the state input of the deep reinforcement learning network to form a state set including the number of queued tasks in the CNC system queue, the number of remaining tasks in the CNC system processing queue, the amount of sub-computing task data, and the amount of computing resource required. Improve the state input of the dual-delay deep deterministic strategy to form a state set including the number of queued tasks in the wireless transmission queue, the number of remaining tasks in the wireless transmission queue, the number of queued tasks in the edge computing queue, and the number of remaining tasks in the edge computing processing queue. S9: Improve the action input of deep reinforcement learning network to form representation The action set of the overall migration action; improve the action input of the dual-delay deep deterministic strategy to form a bandwidth ratio allocation set , computing resource ratio allocation set A collection of actions; S10: Improve the reward value of the deep reinforcement learning network and the double-delay deep deterministic strategy, both defined as the number of successful sub-computation tasks completed in the current time; S11: Randomly initialize the model parameters of the deep reinforcement learning network and the double-delay deep deterministic policy, design an update strategy for the Q-value function of the deep reinforcement learning network, use the experience replay mechanism to store each state transition (state, action, reward, next state) of the deep reinforcement learning network and the double-delay deep deterministic policy in the experience pool, and randomly extract small batches of experience samples from the experience pool for policy training; S12: After completing a specified number of rounds of sample training for the deep reinforcement learning network and the dual-delay deep deterministic strategy, the network strategy is deployed to the CNC system and edge computing facilities of the CNC machine tool to assist the CNC system in real time in completing the dynamic scheduling and optimal allocation of computing tasks.

2. A method for dynamic scheduling of computing tasks of CNC machine tools based on edge computing according to claim 1, characterized in that: In the step S1, the time interval Greater than or equal to 10ms.

3. The method for dynamic scheduling of computing tasks of CNC machine tools based on edge computing according to claim 1, characterized in that: In step S2, the historical monitoring data includes the operating status of the machine tool equipment, fault records and previous task execution status, and the basis for splitting the sub-computing task sequence includes the priority, complexity and required resources of the task.

4. The method for dynamic scheduling of computing tasks of CNC machine tools based on edge computing according to claim 1, characterized in that: In step S3, the sequence analysis of the sub-computing task sequence is based on the task execution time, resource requirements and process requirements to ensure the reasonable scheduling of the sub-computing tasks.

5. The method for dynamic scheduling of computing tasks of CNC machine tools based on edge computing according to claim 1, characterized in that: In step S4, the numerical control system timer is used to accurately record the processing time of the sub-computing task, and the timer resolution is not higher than 10ms.

6. The method for dynamic scheduling of computing tasks of CNC machine tools based on edge computing according to claim 1, characterized in that: In the step S5, the numerical control system timer is used to accurately record the wireless transmission duration of the sub-computing task, and the timer resolution is not higher than 10ms.

7. The method for dynamic scheduling of computing tasks of CNC machine tools based on edge computing according to claim 1, characterized in that: In step S6, the edge computing facility timer is used to accurately record the processing time of the sub-computing task, and the timer resolution is not higher than 10ms.

8. The method for dynamic scheduling of computing tasks of CNC machine tools based on edge computing according to claim 1, characterized in that: In step S7, the dynamic scheduling target takes into account the priority, resource requirements and execution time of the sub-computing tasks.

9. The method for dynamic scheduling of computing tasks of CNC machine tools based on edge computing according to claim 1, characterized in that: In step S9 and step S10, the dual-delay deep deterministic strategy dynamically adjusts the scheduling strategy to adapt to different computing task requirements by updating the state input in real time.

10. The method for dynamic scheduling of computing tasks of CNC machine tools based on edge computing according to claim 1, characterized in that: In the step S11, the extraction strategy of the small batch of experience samples adopts uniform random sampling or priority experience sampling to enhance the learning effect of the scheduling strategy.

Images