Numerical control machining dynamic scheduling method and system for multi-machine cooperation

Abstract

The invention relates to the technical field of numerical control machining production control, in particular to a numerical control machining dynamic scheduling method oriented to multi-machine collaboration, which comprises the following steps of: establishing a directed graph according to real-time production state data, and then obtaining an adjustment priority score of each machining device based on a preset graph neural network according to the directed graph; the plurality of processing devices are scheduled based on the directed graph and the adjustment priority score. Compared with the prior art, the dynamic state of the multi-machine collaborative processing system can be more comprehensively and accurately modeled by constructing the directed graph containing the processing equipment node features and the process relationship, an adjustment priority score calculation mechanism based on the graph neural network is introduced, the adjustment suitability of each equipment in an emergency can be intelligently evaluated, and the processing efficiency of the multi-machine collaborative processing system is improved. The quick response capability of the scheduling scheme to abnormal events is effectively improved, better scheduling performance and higher production efficiency are achieved, and the problem that in the prior art, a multi-machine-tool scheduling method is poor in flexibility is solved.

Description

Technical Field

[0001] The present invention relates to the field of CNC machining production control technology, and in particular to a CNC machining dynamic scheduling method and system for multi-machine collaboration. Background Art

[0002] In automated CNC machining, multi-machine collaboration technology significantly improves efficiency, shortens production cycles, and enhances system flexibility by integrating the processing capabilities of multiple CNC machine tools. Multi-machine collaboration enables parallel task allocation, load balancing, and optimized resource allocation, making it particularly suitable for the efficient machining of complex parts and rapid response to large-scale orders.

[0003] However, existing scheduling methods often encounter the "curse of dimensionality" problem when dealing with multi-machine collaborative scenarios. With the increase in variables such as processing tasks, equipment status, and process constraints, the system must address factors including process sequence constraints, as well as multi-dimensional factors such as equipment, time, and quantity. Furthermore, the often high production volumes drive an exponential increase in decision-making dimensions. This results in traditional optimization algorithms (such as genetic algorithms and particle swarm optimization) being prone to computational time explosion when coordinating multi-machine scheduling, making it difficult to respond promptly and flexibly to unexpected situations in complex production processes.

[0004] Therefore, people need a dynamic scheduling method for CNC machining based on multi-machine collaboration to solve the problem of poor flexibility of existing multi-machine scheduling methods. Summary of the Invention

[0005] Therefore, the present invention provides a dynamic scheduling method and system for CNC machining for multi-machine collaboration, so as to solve the problem of poor flexibility of the multi-machine scheduling method in the prior art.

[0006] The present invention provides a dynamic scheduling method for multi-machine collaboration in numerical control machining, comprising: Acquire real-time production status data and build a directed graph based on it. Each node in the directed graph represents each processing device. Each node also corresponds to a node feature vector that represents the attributes of the processing device. Each edge in the directed graph represents the process relationship existing in the current production. According to the directed graph, an adjustment priority score of each processing equipment is obtained based on a preset graph neural network. The adjustment priority score is used to indicate the suitability of the processing equipment for adjustment in an emergency situation. Obtain real-time production dynamic data; According to real-time production dynamic data, multiple processing equipment are scheduled based on a directed graph and adjusted priority scores.

[0007] In a preferred implementation: the preset graph neural network includes an input layer, multiple aggregation layers, a fully connected layer and an output layer, wherein the input layer is used to input a directed graph, the aggregation layer is used to respectively aggregate the in-degree neighbor nodes and out-degree neighbor nodes of each node, and update the node feature vector of each node, and the fully connected layer is used to respectively input the node feature vector updated by the aggregation layer, and output the adjusted priority score of each node.

[0008] In a preferred implementation, each edge in the directed graph further corresponds to a transportation cost index, which is used to characterize the cost of transporting a workpiece between two processing devices corresponding to the edge; the aggregation layer includes a first weight matrix, a second weight matrix, and a third weight matrix; and the process of updating the node feature vector includes: Calculate the product of the first weight matrix and the node feature vector of the target node to obtain a first aggregation item; Calculate the product of the second weight matrix and the node feature vector of the target node's in-degree neighbor node, and perform weighted averaging on the product based on the transportation cost index to obtain a second aggregation item; Calculate the product of the third weight matrix and the node feature vector of the target node's out-degree neighbor node, and perform weighted averaging on the product based on the transportation cost index to obtain a third aggregation item; An updated node feature vector of the target node is obtained according to the sum of the first aggregation item, the second aggregation item, and the third aggregation item.

[0009] In a preferred implementation, the training data of the preset graph neural network is obtained by: Obtain historical adjustment logs, and obtain historical adjustment devices and quantitative values ​​of adjustment results corresponding to the historical adjustment devices based on the historical adjustment logs; According to the quantified value of the adjustment result, the real label corresponding to the historical adjustment device is established to obtain the training data.

[0010] In a preferred implementation, the training data of the preset graph neural network is obtained by: Obtain historical status statistics of processing equipment; Based on historical state statistical data, pseudo labels of processing equipment are generated, and the pseudo labels are used as the true labels of the processing equipment to obtain training data.

[0011] In a preferred implementation, scheduling multiple processing equipment based on real-time production dynamic data, a directed graph, and adjusted priority scores includes: According to the real-time production dynamic data, the faulty equipment is found; According to the directed graph, obtain the sibling nodes of the node corresponding to the faulty device, as well as the downstream nodes of the sibling nodes; Obtaining an alternative path based on the sibling node and its downstream nodes according to the adjusted priority scores, wherein the sum of the adjusted priority scores of all nodes in the alternative path is the smallest; Scheduling multiple processing equipment based on alternative paths.

[0012] In a preferred implementation, scheduling multiple processing equipment based on real-time production dynamic data, a directed graph, and adjusted priority scores includes: Get insertion order data based on real-time production dynamic data; According to the process corresponding to the workpiece in the insertion order data, based on the directed graph and the adjustment priority score, the insertion order path is obtained, where the sum of the adjustment priority scores of all nodes in the insertion order path is the smallest; Based on the insertion order path, multiple processing equipment are scheduled.

[0013] The present invention also provides a dynamic scheduling system for multi-machine collaboration in numerical control machining, comprising: The initialization module is used to obtain real-time production status data and build a directed graph based on the real-time production status data. Each node in the directed graph represents each processing equipment. Each node also corresponds to a node feature vector that represents the attributes of the processing equipment. Each edge in the directed graph represents the process relationship existing in the current production; A priority analysis module is used to obtain an adjustment priority score for each processing equipment based on a directed graph and a preset graph neural network. The adjustment priority score is used to indicate the suitability of the processing equipment for adjustment in an emergency situation. Production monitoring module, used to obtain real-time production dynamic data; The production scheduling module is used to schedule multiple processing equipment based on real-time production dynamic data, a directed graph and adjusted priority scores.

[0014] The present invention further provides an electronic device, comprising: memory and processor; The memory is used to store the program, and the processor is used to execute the steps of any one of the above-mentioned methods for dynamic scheduling of CNC machining for multi-machine collaboration when executing the program.

[0015] The present invention also provides a computer-readable storage medium for storing computer-readable programs or instructions. When the program or instructions are executed by a processor, the steps in any of the above-mentioned dynamic scheduling methods for CNC machining for multi-machine collaboration can be implemented.

[0016] The beneficial effects of adopting the above embodiment are: The present invention provides a dynamic scheduling method for CNC machining for multi-machine collaboration. The method first obtains real-time production status data and establishes a directed graph based on the real-time production status data. Each node in the directed graph represents each processing equipment. Each node also corresponds to a node feature vector that characterizes the attributes of the processing equipment. Each edge in the directed graph represents the process relationship existing in the current production. Then, based on the directed graph and a preset graph neural network, an adjustment priority score is obtained for each processing equipment. The adjustment priority score is used to indicate the suitability of the processing equipment for adjustment in an emergency situation. Then, the real-time production dynamic data is obtained and multiple processing equipment are scheduled based on the directed graph and the adjustment priority score. Compared with the existing technology, the present invention can more comprehensively and accurately model the dynamic state of a multi-machine collaborative processing system by constructing a directed graph containing the node characteristics of processing equipment and the process relationship, so that scheduling decisions can fully consider the equipment attributes and process dependencies, and introduce an adjustment priority score calculation mechanism based on a graph neural network, which can intelligently evaluate the adjustment suitability of each device in an emergency situation, effectively improving the scheduling plan's ability to respond quickly to abnormal events. Through the combination of graph neural networks and dynamic data driving, it effectively alleviates the "dimensionality disaster" problem faced by traditional scheduling methods, greatly improves the efficiency of solving large-scale multi-machine collaborative scheduling problems, enables the system to adjust the scheduling strategy in time according to the latest production status, significantly improves the real-time and adaptability of scheduling, achieves better scheduling performance and higher production efficiency, and solves the problem of poor flexibility of multi-machine scheduling methods in the existing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A method flow chart of the dynamic scheduling method for multi-machine collaboration in numerical control machining provided by the present invention; Figure 2 for Figure 1 Specific step diagram of step S104; Figure 3 for Figure 1 Another specific step diagram of step S104; Figure 4 This is a system architecture diagram of the dynamic scheduling system for CNC machining with multiple machines provided by the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] See also Figure 1A specific embodiment of the present invention discloses a dynamic scheduling method for multi-machine collaboration in numerical control machining, comprising: S101. Acquire real-time production status data and establish a directed graph based on the real-time production status data. Each node in the directed graph represents each processing equipment. Each node also corresponds to a node feature vector representing the attributes of the processing equipment. Each edge in the directed graph represents the process relationship existing in the current production. S102. Obtaining an adjustment priority score for each processing equipment based on the directed graph and a preset graph neural network, where the adjustment priority score is used to indicate the suitability of the processing equipment for adjustment in an emergency situation; S103, obtaining real-time production dynamic data; S104. Schedule multiple processing equipment according to real-time production dynamic data, based on the directed graph and adjusted priority scores.

[0020] Real-time production status data refers to any data that can be obtained during the production process, such as the operating status of machine tools, current production progress, order status, etc. Real-time production dynamic data refers to data that changes during the current production process, such as equipment failures and urgent orders.

[0021] Each node in a directed graph corresponds to a piece of processing equipment, such as a CNC lathe or multi-axis machining center. The node feature vector corresponding to each node includes the equipment's attributes that must be considered during dynamic scheduling, such as the equipment type, current processing workload, remaining process time, equipment load, fault status (normal / downtime / maintenance), and available time windows (e.g., tool change / maintenance periods). The edges in the directed graph reflect the process sequence at the time of production. Understandably, since most CNC machining equipment utilizes a subtractive manufacturing model, the directed graph typically constructed is a directed acyclic graph.

[0022] Compared to the complex multi-factor optimization algorithms of existing technologies, this invention constructs the production state as a directed graph, which can more comprehensively and accurately model the dynamic state of a multi-machine collaborative processing system, allowing scheduling decisions to fully consider equipment attributes and process dependencies. At the same time, based on the graph neural network's adjustment priority score calculation mechanism, the complex multiple factors in traditional methods are summarized into a single adjustment priority score. Subsequent dynamic adjustments can be made based solely on the adjustment priority score, avoiding the problem of "dimensionality explosion" and greatly improving response speed, enabling the production process to quickly adapt to complex and changing production conditions.

[0023] Specifically, in a preferred embodiment, the preset graph neural network includes an input layer, multiple aggregation layers, a fully connected layer and an output layer, wherein the input layer is used to input a directed graph, the aggregation layer is used to respectively aggregate the in-degree neighbor nodes and out-degree neighbor nodes of each node, and update the node feature vector of each node, and the fully connected layer is used to respectively input the node feature vectors updated by the aggregation layer, and output the adjusted priority score of each node.

[0024] Because this embodiment analyzes a directed graph, it improves upon existing graph neural network technology to account for process sequence constraints within the production process. Specifically, the aggregation layer in this embodiment divides neighbor nodes into in-degree neighbor nodes (device nodes corresponding to the previous process of the current node) and out-degree neighbor nodes (device nodes corresponding to the next process of the current node) when aggregating node information. These nodes are then aggregated separately, achieving explicit modeling of process sequence constraints and enabling scheduling decisions to fully consider the logical dependencies between processes. This direction-aware aggregation mechanism not only more accurately captures the process correlation characteristics between devices but also effectively avoids scheduling conflicts caused by ignoring process sequence in traditional methods. Through the feature update mechanism of the multi-layer aggregation structure, the network can gradually integrate multi-hop neighbor information, thereby more comprehensively evaluating the adjustment suitability of each device in a dynamic environment.

[0025] Furthermore, in a preferred embodiment, each edge in the directed graph further corresponds to a transportation cost index, which is used to characterize the cost of transporting a workpiece between two processing devices corresponding to the edge; the aggregation layer includes a first weight matrix, a second weight matrix, and a third weight matrix; and the process of updating the node feature vector includes: Calculate the product of the first weight matrix and the node feature vector of the target node to obtain a first aggregation item; Calculate the product of the second weight matrix and the node feature vector of the target node's in-degree neighbor node, and perform weighted averaging on the product based on the transportation cost index to obtain a second aggregation item; Calculate the product of the third weight matrix and the node feature vector of the target node's out-degree neighbor node, and perform weighted averaging on the product based on the transportation cost index to obtain a third aggregation item; An updated node feature vector of the target node is obtained according to the sum of the first aggregation item, the second aggregation item, and the third aggregation item.

[0026] In practice, the operating conditions of auxiliary equipment (such as AGV carts, manipulators, and other equipment used to transport parts between different processing equipment), the physical distance between two processing equipment, and the sorting strategy during workpiece diversion will also affect the efficiency of production and processing (whether waiting is required for part transfer). Therefore, this embodiment further introduces a transportation cost index to represent this influencing factor in order to control the impact ratio of different auxiliary equipment transportation methods during aggregation.

[0027] The transportation cost index can be defined artificially based on experience or experiments. For example, the transportation cost index of conveyor belt transportation is recorded as 1, the transportation cost index of automated equipment handling is recorded as 2, the transportation cost index of manual sorting and handling of workpieces is recorded as 3, and the transportation cost index of long physical distances requiring forklift transportation is recorded as 4, etc.

[0028] This embodiment quantitatively characterizes the actual cost of transferring workpieces between devices by embedding the transportation cost index in the edge attributes of the directed graph, so that scheduling decisions can more realistically reflect the constraints of the production site. In terms of network structure design, by adopting a differentiated aggregation method of a three-weight matrix and weighted fusion of in-degree and out-degree neighbor information based on the transportation cost index, the collaborative optimization of process dependencies and logistics transportation costs is achieved. This design can not only avoid the problem of infeasibility of scheduling plans caused by ignoring transportation factors in traditional methods, but also effectively reduce the idle time of equipment caused by parts waiting for transportation. The solution of this embodiment can significantly reduce the transportation cost of the overall scheduling plan, improve equipment utilization, and automatically capture the complex relationship between transportation constraints and process requirements through end-to-end feature learning, providing a more accurate and reliable solution for efficient collaborative scheduling in complex manufacturing environments.

[0029] The formula for the above process is: ; in, Represents the sequence number of the node in the directed graph, Representation node The set of in-degree neighbor nodes of Represents the sequence number of the node in the set of in-degree neighbor nodes, Representation node The set of out-degree neighbor nodes, Representation node The sequence number of the node in the set of out-degree neighbor nodes, Indicates the number of layers of the aggregation layer, Representation node In the The node feature vector input when the aggregation layer is Representation node In the The node feature vector input when the aggregation layer is In the The node feature vector updated by the aggregation layer, Representation node In the The node feature vector input when the aggregation layer is Representation node In the The node feature vector input when the aggregation layer is Representation node and nodes The transport cost index of the edge between Representation node and nodes The transport cost index of the edge between 、 and They are the first weight matrix, the second weight matrix and the third weight matrix, all of which are the targets of the preset graph neural network training and are obtained by training. It is a preset activation function (such as softmax, relu, etc.). The three items in the activation function are the first aggregation item, the second aggregation item, and the third aggregation item.

[0030] It is understandable that the "adjustment priority score" output by the preset graph neural network in the previous article is not an actual physical quantity, so there may be a problem of difficulty in preparing training data, that is, samples that directly mark "whether the node is suitable for adjustment" (such as "total delay reduction after adjustment") are difficult to obtain on a large scale in actual production. Therefore, the present invention also provides a preferred embodiment to quantify the "adjustment priority score" to solve the problem of difficulty in preparing training data.

[0031] In one embodiment, the training data of the preset graph neural network is obtained by: Obtain historical adjustment logs, and obtain historical adjustment devices and quantitative values ​​of adjustment results corresponding to the historical adjustment devices based on the historical adjustment logs; According to the quantified value of the adjustment result, the real label corresponding to the historical adjustment device is established to obtain the training data.

[0032] This example cleverly leverages the factory's existing historical adjustment log data. By mining historical adjustment records and their corresponding actual production results (such as quantifiable indicators like total delay reduction after adjustment and average equipment load rate), an objective and realistic training sample labeling system is established. This approach overcomes the limitations of traditional supervised learning, which requires manual data labeling. This not only significantly reduces data preparation costs but also ensures the authenticity of training samples and their adaptability to production environments. Furthermore, it converts the abstract "adjustment priority score" into a quantifiable historical adjustment effect assessment, enabling continuous model optimization based on real-world production feedback and effectively avoiding the disconnect between simulated data and actual scenarios.

[0033] It can be understood that in the above process, true labels are established for devices with adjustment records. For those devices without adjustment records (such as newly purchased equipment), a default value can be set as the true label, and a "directed graph-true label mapping of each node" can be established as input and output data to form a training set.

[0034] Furthermore, the present invention further provides an embodiment in which a more scientific and reasonable true label is established for a device with no adjustment record: In this embodiment, the training data of the preset graph neural network is obtained by: Obtain historical status statistics of processing equipment; Based on historical state statistical data, pseudo labels of processing equipment are generated, and the pseudo labels are used as the true labels of the processing equipment to obtain training data.

[0035] This embodiment uses historical equipment status statistics (such as equipment failure frequency, average load rate, process completion time, etc.) as a basis. Through scientific statistical analysis and feature engineering, it generates pseudo-labels that reflect the suitability of equipment adjustments (i.e., performing relevant operations on the above historical status statistics to obtain specific numerical values. For example, if the historical average load of processing equipment is greater than 90%, the pseudo-label for that equipment can be set to 0.9, indicating a high adjustment priority). This not only establishes reasonable initial training labels for new equipment but also dynamically reflects the changing trends of equipment performance. This method uses the fusion analysis of multi-dimensional status data to enable pseudo-labels to more accurately reflect the actual adjustment needs of the equipment. As equipment operation data continues to accumulate, the pseudo-labels can be adaptively updated to ensure continuous model optimization, avoid the subjectivity of manual labeling, and make the labels more objective and consistent.

[0036] Further, combined Figure 2 As shown, in a preferred embodiment, the above step S104, scheduling multiple processing equipment according to real-time production dynamic data, based on a directed graph and adjusting priority scores, specifically includes: S201. Obtain faulty equipment based on real-time production dynamic data; S202: Obtain, based on the directed graph, the sibling nodes of the node corresponding to the faulty device and the downstream nodes of the sibling nodes; S203: Obtain an alternative path based on the adjusted priority scores, the sibling node, and the downstream nodes of the sibling node, wherein the sum of the adjusted priority scores of all nodes in the alternative path is the smallest; S204: Scheduling multiple processing equipment based on alternative paths.

[0037] The above process is a scheduling method for the event of an equipment failure. During scheduling, an adjustment plan can be quickly obtained based solely on the structure of the directed graph and the adjustment priority score. Specifically, this embodiment is entirely based on preprocessed structured data and model calculation results, avoiding real-time optimization calculation delays when a failure occurs and significantly shortening scheduling decision-making time. At the same time, graph structure analysis ensures the process feasibility of alternative paths, effectively avoiding the risk of violating process constraints, and the priority score ensures the comprehensive optimality of the selected path in terms of multiple objectives such as equipment adjustment cost and load balancing.

[0038] Further, combined Figure 3 As shown, in a preferred embodiment, the above step S104, scheduling multiple processing equipment according to real-time production dynamic data based on a directed graph and adjusted priority scores, further includes: S301. Obtain insertion order data based on real-time production dynamic data; S302: Obtain an insertion path based on the directed graph and the adjustment priority scores according to the process corresponding to the workpiece in the insertion data, wherein the sum of the adjustment priority scores of all nodes in the insertion path is the smallest; S303: Scheduling multiple processing equipment based on the insertion order path.

[0039] The above process is a scheduling method for emergency insertion orders. When an emergency insertion order is received, the system can immediately analyze the workpiece process route and identify the feasible insertion order path with the lowest adjustment cost from a global perspective. The path selection comprehensively considers multiple factors such as the current load status of the equipment, the process connection relationship, and the adjustment priority. Similarly, this embodiment also ensures the strict compliance of the insertion order path in terms of process feasibility based on the global search of the graph structure, avoids the risk of process conflict, and realizes multi-objective optimization by introducing the adjustment priority score. While meeting the needs of emergency orders, it minimizes the impact on the original production plan. The pre-calculation and structured query mechanism greatly shortens the scheduling decision time, enabling the system to respond quickly.

[0040] It is understandable that, knowing the adjusted priority score of each node, how to find the alternative path and the insertion path with the smallest total adjusted priority score is an existing technology that can be understood by those skilled in the art and can be implemented using any existing mathematical algorithm. Therefore, this article will not elaborate on it.

[0041] Combine Figure 4 As shown, the present invention also provides a dynamic scheduling system for CNC machining for multi-machine collaboration, comprising: Initialization module 410 is used to obtain real-time production status data and build a directed graph based on the real-time production status data. Each node in the directed graph represents each processing equipment. Each node also corresponds to a node feature vector representing the attributes of the processing equipment. Each edge in the directed graph represents the process relationship existing in the current production; Priority analysis module 420, configured to obtain an adjustment priority score for each processing equipment based on the directed graph and a preset graph neural network, wherein the adjustment priority score is used to indicate the suitability of the processing equipment for adjustment in an emergency situation; Production monitoring module 430, used to obtain real-time production dynamic data; The production scheduling module 440 is used to schedule multiple processing equipment according to real-time production dynamic data based on a directed graph and adjusted priority scores.

[0042] It should be noted here that the corresponding system provided in the above embodiments is a computer program product, which can implement the technical solutions described in the above method embodiments. The specific implementation principles of the above modules or units can be found in the corresponding contents in the above method embodiments, which will not be repeated here.

[0043] The present invention further provides an electronic device, comprising: memory and processor; The memory is used to store the program, and the processor is used to execute the steps of any one of the above-mentioned methods for dynamic scheduling of CNC machining for multi-machine collaboration when executing the program.

[0044] The present invention also provides a computer-readable storage medium for storing computer-readable programs or instructions. When the program or instructions are executed by a processor, the steps in any of the above-mentioned dynamic scheduling methods for CNC machining for multi-machine collaboration can be implemented.

[0045] The present invention provides a dynamic scheduling method for CNC machining for multi-machine collaboration. The method first obtains real-time production status data and establishes a directed graph based on the real-time production status data. Each node in the directed graph represents each processing equipment. Each node also corresponds to a node feature vector that characterizes the attributes of the processing equipment. Each edge in the directed graph represents the process relationship existing in the current production. Then, based on the directed graph and a preset graph neural network, an adjustment priority score is obtained for each processing equipment. The adjustment priority score is used to indicate the suitability of the processing equipment for adjustment in an emergency situation. Then, the real-time production dynamic data is obtained and multiple processing equipment are scheduled based on the directed graph and the adjustment priority score. Compared with the existing technology, the present invention can more comprehensively and accurately model the dynamic state of a multi-machine collaborative processing system by constructing a directed graph containing the node characteristics of processing equipment and the process relationship, so that scheduling decisions can fully consider the equipment attributes and process dependencies, and introduce an adjustment priority score calculation mechanism based on a graph neural network, which can intelligently evaluate the adjustment suitability of each device in an emergency situation, effectively improving the scheduling plan's ability to respond quickly to abnormal events. Through the combination of graph neural networks and dynamic data driving, it effectively alleviates the "dimensionality disaster" problem faced by traditional scheduling methods, greatly improves the efficiency of solving large-scale multi-machine collaborative scheduling problems, enables the system to adjust the scheduling strategy in time according to the latest production status, significantly improves the real-time and adaptability of scheduling, achieves better scheduling performance and higher production efficiency, and solves the problem of poor flexibility of multi-machine scheduling methods in the existing technology.

[0046] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referenced to each other.

[0047] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A dynamic scheduling method for CNC machining based on multi-machine collaboration, characterized in that: include: Acquire real-time production status data and build a directed graph based on it. Each node in the directed graph represents each processing device. Each node also corresponds to a node feature vector that represents the attributes of the processing device. Each edge in the directed graph represents the process relationship existing in the current production. According to the directed graph, an adjustment priority score of each processing equipment is obtained based on a preset graph neural network. The adjustment priority score is used to indicate the suitability of the processing equipment for adjustment in an emergency situation. Obtain real-time production dynamic data; According to real-time production dynamic data, multiple processing equipment are scheduled based on a directed graph and adjusted priority scores.

2. The dynamic scheduling method for multi-machine collaboration in numerical control machining according to claim 1, characterized in that: The preset graph neural network includes an input layer, multiple aggregation layers, a fully connected layer and an output layer. The input layer is used to input a directed graph, the aggregation layer is used to aggregate the in-degree neighbor nodes and out-degree neighbor nodes of each node respectively, and update the node feature vector of each node. The fully connected layer is used to input the node feature vectors updated by the aggregation layer respectively, and output the adjusted priority score of each node.

3. The dynamic scheduling method for multi-machine collaboration in numerical control machining according to claim 2, characterized in that: Each edge in the directed graph also corresponds to a transportation cost index, which is used to represent the cost of transporting the workpiece between the two processing equipment corresponding to the edge; The aggregation layer includes a first weight matrix, a second weight matrix, and a third weight matrix; The process of updating the node feature vector includes: Calculate the product of the first weight matrix and the node feature vector of the target node to obtain a first aggregation item; Calculate the product of the second weight matrix and the node feature vector of the target node's in-degree neighbor node, and perform weighted averaging on the product based on the transportation cost index to obtain a second aggregation item; Calculate the product of the third weight matrix and the node feature vector of the target node's out-degree neighbor node, and perform weighted averaging on the product based on the transportation cost index to obtain a third aggregation item; An updated node feature vector of the target node is obtained according to the sum of the first aggregation item, the second aggregation item, and the third aggregation item.

4. The dynamic scheduling method for multi-machine collaboration in numerical control machining according to claim 1, characterized in that: The training data of the preset graph neural network is obtained in the following way: Obtain historical adjustment logs, and obtain historical adjustment devices and quantitative values ​​of adjustment results corresponding to the historical adjustment devices based on the historical adjustment logs; According to the quantified value of the adjustment result, the real label corresponding to the historical adjustment device is established to obtain the training data.

5. The dynamic scheduling method for multi-machine collaboration in numerical control machining according to claim 1, characterized in that: The training data of the preset graph neural network is obtained in the following way: Obtain historical status statistics of processing equipment; Based on historical state statistical data, pseudo labels of processing equipment are generated, and the pseudo labels are used as the true labels of the processing equipment to obtain training data.

6. The dynamic scheduling method for multi-machine collaboration in numerical control machining according to claim 1, characterized in that: Based on real-time production dynamics data, multiple processing equipment are scheduled based on a directed graph and adjusted priority scores, including: According to the real-time production dynamic data, the faulty equipment is found; According to the directed graph, obtain the sibling nodes of the node corresponding to the faulty device, as well as the downstream nodes of the sibling nodes; Obtaining an alternative path based on the sibling node and its downstream nodes according to the adjusted priority scores, wherein the sum of the adjusted priority scores of all nodes in the alternative path is the smallest; Scheduling multiple processing equipment based on alternative paths.

7. The dynamic scheduling method for multi-machine collaboration in numerical control machining according to claim 1, characterized in that: Based on real-time production dynamics data, multiple processing equipment are scheduled based on a directed graph and adjusted priority scores, including: Get insertion order data based on real-time production dynamic data; According to the process corresponding to the workpiece in the insertion order data, based on the directed graph and the adjustment priority score, the insertion order path is obtained, where the sum of the adjustment priority scores of all nodes in the insertion order path is the smallest; Based on the insertion order path, multiple processing equipment are scheduled.

8. A dynamic scheduling system for CNC machining for multi-machine collaboration, characterized in that: include: The initialization module is used to obtain real-time production status data and build a directed graph based on the real-time production status data. Each node in the directed graph represents each processing equipment. Each node also corresponds to a node feature vector that represents the attributes of the processing equipment. Each edge in the directed graph represents the process relationship existing in the current production; A priority analysis module is used to obtain an adjustment priority score for each processing equipment based on a directed graph and a preset graph neural network. The adjustment priority score is used to indicate the suitability of the processing equipment for adjustment in an emergency situation. Production monitoring module, used to obtain real-time production dynamic data; The production scheduling module is used to schedule multiple processing equipment based on real-time production dynamic data, a directed graph and adjusted priority scores.

9. An electronic device, characterized in that: include: memory and processor; The memory is used to store a program, and the processor is used to execute the steps of the dynamic scheduling method for CNC machining for multi-machine collaboration described in any one of claims 1 to 7 when executing the program.

10. A computer-readable storage medium, characterized in that Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps of the dynamic scheduling method for CNC machining for multi-machine collaboration according to any one of claims 1 to 7.

Images