Distributed control machining method based on CNC fine carving machining task

Abstract

The present application relates to the technical field of fine carving processing, and particularly relates to a distributed control processing method based on CNC fine carving processing tasks, which first acquires multiple distributed CNC devices in a region and constructs a topological network through an interactive scheduling platform; secondly, task data sets are extracted based on historical fine carving records, and the topological network is divided into multiple co-state classes according to the task data sets; subsequently, a distributed collaborative processing controller containing multiple strategies is constructed in combination with the data sets and the co-state results; finally, real-time processing state data is collected by activating the topological network to input the controller, and distributed collaborative control of multiple devices is realized. The present application identifies task commonality by mining historical data rules, dynamically matches optimal control strategies, effectively solves the problems of poor device collaboration, uneven load and weak adaptability in traditional distributed processing, and significantly improves the processing consistency, production efficiency and flexibility level of the system.

Description

Technical Field

[0001] This invention relates to the field of precision carving technology, and in particular to a distributed control machining method based on CNC precision carving tasks. Background Technology

[0002] With the rapid development of industries such as high-end manufacturing, precision molds, consumer electronics, and semiconductor packaging, the demand for high-precision machining of complex curved surfaces and microstructures is increasing. CNC (Computer Numerical Control) engraving technology, due to its high precision, high flexibility, and good repeatability, has become one of the core processes in modern precision manufacturing. In actual production scenarios, to improve processing efficiency and system reliability, a distributed machining system composed of multiple CNC engraving machines is often used to achieve parallel task processing and resource collaborative scheduling.

[0003] However, most current distributed CNC machining systems still employ centralized task allocation or independent control modes based on fixed rules, lacking effective information exchange and collaborative decision-making mechanisms between devices. Such methods struggle to adapt to varying machining task characteristics, such as material type, tool wear, and machining path complexity, easily leading to uneven equipment load, poor machining consistency, and low overall efficiency. Especially when handling high-precision, batch, and multi-product mixed-line production tasks, traditional control strategies struggle to dynamically respond to real-time machining status changes and lack effective utilization of historical machining experience, limiting the adaptive and optimization capabilities of intelligent manufacturing systems.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this invention is to provide a distributed control machining method based on CNC precision carving tasks. This method aims to solve the technical problems of existing distributed CNC precision carving systems, which lack a task commonality identification mechanism based on in-depth mining of historical data and intelligent collaborative decision-making capabilities of multiple devices. As a result, these systems are unable to adapt to complex and ever-changing machining conditions, leading to uneven equipment load, poor machining consistency, and low overall efficiency.

[0006] To achieve the above objectives, the present invention provides a distributed control machining method based on CNC precision carving tasks, the method comprising: An interactive machining task scheduling platform acquires multiple distributed CNC precision carving machines in the target machining task area and constructs a machining equipment topology network. Based on the historical fine carving records of the target processing task area, extract the fine carving task dataset; Based on the processing task dataset, the processing equipment topology network is partitioned into task common modes, and the common mode partitioning results are obtained, wherein the common mode partitioning results include multiple task common mode partitioning classes; A distributed collaborative processing controller is constructed using the processing task dataset and the comorphic partitioning results, wherein the distributed collaborative processing controller includes multiple collaborative processing control strategies; The processing equipment topology network is activated to collect real-time processing status data of the target processing task area, which is then input into the distributed collaborative processing controller to perform distributed collaborative control processing of multiple distributed CNC precision carving processing equipment.

[0007] Optionally, the interactive machining task scheduling platform acquires multiple distributed CNC engraving machines in the target machining task area and constructs a machining machine topology network, including: Establish a communication connection with the processing task scheduling platform to obtain the target processing task configuration file; Based on the target machining task configuration file, the location information, machining parameter information and current status information of multiple distributed CNC precision carving machines within the target machining task area are extracted and stored as basic information of the machining machines. Analyze and determine the connection relationships and communication methods of multiple distributed CNC precision carving machines within the target machining task area, and construct a distributed machining equipment network; The basic information of the processing equipment is updated to the distributed processing equipment network to obtain the processing equipment topology network.

[0008] Optionally, the step of performing task comorphic partitioning of the processing equipment topology network based on the processing task dataset and obtaining the comorphic partitioning result includes: The processing task dataset is preprocessed to obtain a standard processing task dataset; Initialize the task common-mode partitioning algorithm and input the standard processing task dataset for common-mode partitioning to obtain a classified processing dataset, wherein the classified processing dataset includes multiple processing data groups; Based on the association between multiple processing data groups and multiple task comorphic partitioning classes, multiple task comorphic partitioning classes are mapped and established, and stored as the comorphic partitioning results.

[0009] Optionally, before performing task comorphic partitioning of the processing equipment topology network based on the processing task dataset, the method further includes: Analyze the processing task dataset to obtain the key feature set for task commonality partitioning; The key feature set is divided according to the commonality of the task, and the dimensionality reduction and size reduction processing of the processing task dataset is performed.

[0010] Optionally, the step of analyzing the processing task dataset to obtain a set of key features for task commonality partitioning includes: Based on the principle of standardization, the dataset of the processing tasks is decentralized. Based on the decentralized results, a standard processing data matrix is ​​constructed with the number of processing tasks in the processing task dataset as the number of rows in the matrix and the number of processing index categories in the processing task dataset as the number of columns in the matrix. Establish the correlation matrix of the standard processing data matrix, calculate the eigenvalues ​​and eigenvectors of the correlation matrix, and obtain the feature set of the correlation matrix; Based on the variance contribution scores of the feature set of the correlation matrix, a cumulative variance contribution sequence is obtained by accumulating the variance contribution scores, wherein the cumulative variance contribution sequence includes multiple cumulative variance contribution scores, and the cumulative variance contribution scores have processing feature class labels. Based on the preset contribution rate constraint, the cumulative variance contribution score is filtered according to the cumulative variance contribution sequence, and the processing feature class label of the filtering result is extracted to generate the task commonality partitioning key feature set.

[0011] Optionally, constructing a distributed collaborative processing controller using the processing task dataset and the comorphic partitioning result includes: Randomly select from the common-state partitioning results to obtain the first task common-state partitioning class; Based on the first task common-state classification, and based on the common-state classification result, the first processing data group is invoked; Based on the first set of processed data, a first collaborative processing control strategy is obtained through supervised training. The first collaborative processing control strategy is stored in the distributed collaborative processing controller.

[0012] Optionally, the distributed collaborative control of multiple distributed CNC engraving machines includes: Analyze the real-time processing status data to determine the common-state classification of real-time tasks; Based on the real-time task common-state classification, the multiple collaborative processing control strategies in the distributed collaborative processing controller are traversed for control matching. Based on the control matching results, the distributed collaborative processing controller is activated to make processing control decisions. The processing equipment topology network distributes the processing control decision results to multiple distributed CNC precision carving processing equipment for distributed collaborative control processing.

[0013] Optionally, the distributed collaborative control process of multiple distributed CNC engraving machines also includes: Periodically collect data on actual machining dimensions and tool wear from each distributed CNC precision carving equipment, and summarize them to generate a real-time machining error dataset; Calculate the average overall processing error under the current real-time task common-state partitioning class. When the average overall processing error exceeds the preset processing accuracy threshold, re-partition the current processing state data into common-state partitions, re-match the collaborative processing control strategy, and update and issue processing control commands to complete the dynamic correction of processing control.

[0014] Furthermore, to achieve the above objectives, the present invention also provides a distributed control machining device based on CNC precision carving tasks. The device includes: a memory, a processor, and a distributed control machining program based on CNC precision carving tasks stored in the memory and executable on the processor. The distributed control machining program based on CNC precision carving tasks is configured to implement the steps of the distributed control machining method based on CNC precision carving tasks as described above.

[0015] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium storing a distributed control machining program based on a CNC engraving task. When the distributed control machining program based on the CNC engraving task is executed by a processor, it implements the steps of the distributed control machining method based on a CNC engraving task as described above.

[0016] This invention provides a distributed control machining method based on CNC precision carving tasks. By introducing a task common-state partitioning mechanism, this method can deeply mine potential patterns in historical machining data, classifying tasks with similar machining characteristics (such as material properties, tool path complexity, and precision requirements). This allows the system to match the optimal collaborative control strategy for each type of common-state task, effectively avoiding machining quality fluctuations caused by traditional one-size-fits-all control, and significantly improving the consistency of machining results and overall precision stability among multiple machines. The constructed machining equipment topology network, combined with a distributed collaborative machining controller, breaks through the limitations of traditional single-machine independent control or simple centralized scheduling. The system can dynamically adjust the operating parameters and task allocation of each machine based on real-time collected machining status data, automatically balancing the load within the cluster and preventing some machines from being overloaded while others are idle, thereby maximizing the utilization of equipment resources, shortening the overall production cycle, and improving production efficiency. This method integrates historical data modeling and real-time status feedback, giving the machining system powerful self-learning and adaptive capabilities. Faced with complex production scenarios involving multiple varieties, small batches, or frequent process changes, the system can quickly switch or generate new collaborative control strategies based on common-state partitioning results without the need for tedious manual reprogramming. This significantly improves the production line's responsiveness to changes in market demand and its flexible manufacturing capabilities. By extracting datasets from historical precision carving records and constructing control strategies, it effectively digitizes and models past successful processing experiences. This not only reduces reliance on highly skilled operators but also lowers scrap rates and unplanned equipment downtime by predicting and avoiding common processing anomalies such as excessive tool wear and excessive vibration, thus effectively reducing long-term maintenance and trial-and-error costs. The method constructs a complete closed loop from data acquisition, feature extraction, common-state analysis to collaborative control, enabling data-driven decision-making in the processing process. This not only solves the problems of information silos and poor collaboration in current distributed processing systems but also lays a solid technical foundation for subsequent integration with higher-level industrial internet platforms, predictive maintenance, and intelligent management of the entire process. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating an embodiment of the distributed control machining method based on CNC precision carving tasks according to the present invention. Figure 2 This is a flowchart illustrating another embodiment of the distributed control machining method based on CNC precision carving tasks according to the present invention.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] Reference Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the distributed control machining method based on CNC precision carving tasks according to the present invention.

[0021] In one embodiment, the distributed control machining method based on CNC precision carving tasks includes: Step S100: The interactive processing task scheduling platform acquires multiple distributed CNC precision carving processing devices in the target processing task area and constructs a processing device topology network.

[0022] The interactive machining task scheduling platform can be a task distribution interface system for receiving machining task instructions and communicating with distributed CNC equipment. It can serve as a task input entry point, coordinating the initial mapping relationship between the target machining task area and equipment resources. In this embodiment, the interactive machining task scheduling platform can interface with the upper-layer MES or local scheduling system through industrial communication protocols (such as OPCUA, ModbusTCP). The target machining task area can be the physical or logical production unit range to which the currently pending machining task belongs, and can be used to define the spatial boundaries for historical data extraction and equipment network construction. Multiple distributed CNC engraving machines can be a cluster of engraving machine tools deployed within the same target machining task area, possessing independent CNC systems but capable of network communication. They can be used as physical execution units to perform specific engraving operations. The machining equipment topology network can be a logical network structure composed of multiple distributed CNC engraving machines and their communication, and can be used to support the exchange of status information and the transmission of collaborative control instructions between devices. For example, the machining equipment topology network can establish an adjacency matrix or graph structure based on device IP addresses, physical locations, or process dependencies. Furthermore, the processing equipment topology network can collaborate with the distributed collaborative processing controller to provide equipment status reporting channels and control command distribution paths.

[0023] The interactive machining task scheduling platform acquires information on multiple distributed CNC engraving machines within the target machining task area. This can be achieved by querying the list of available CNC machines and their status information within the target area through the scheduling platform. Furthermore, this operation can be accomplished by calling a RESTful API to obtain the machine list from the cloud scheduling center, or by broadcasting on a local area network to probe for responding machines and establish connections, thereby establishing the initial association between tasks and machine resources. Constructing the machining machine topology network can be achieved by creating a logical graph between the machines based on the acquired machine list and their communication capabilities. In an exemplary embodiment, this operation can be implemented by constructing a nearest-neighbor communication topology based on the physical location of the machines, or by constructing a directed task flow topology according to the process flow sequence, thereby forming a network foundation that supports information interaction and collaborative control.

[0024] Step S200: Extract the fine carving task dataset based on the historical fine carving records of the target processing task area.

[0025] The historical engraving records for the target processing task area can be a collection of original operation logs and result data for all engraving tasks completed within the target processing task area, which can be used as the original basis for constructing the engraving task dataset. In this embodiment, the historical engraving records for the target processing task area can be exported from local device storage, a database, or a Manufacturing Execution System (MES). The engraving task dataset can be a structured data set containing task features and processing results extracted from the historical engraving records, which can be used to support task commonality partitioning and control strategy modeling. For example, the engraving task dataset can be obtained by standardizing the original records through data cleaning and feature engineering. In a specific embodiment, the engraving task dataset can include, but is not limited to, one or more of the following: material property-dominated datasets, path complexity-dominated datasets, and accuracy requirement-dominated datasets.

[0026] Based on historical fine carving records of the target processing task area, a fine carving task dataset is extracted. This can be achieved by extracting features from historical records and structuring them to form a labeled data table. Furthermore, this operation can be implemented by using ETL tools to extract and transform fields (such as material type, number of path points, and surface roughness) from the database, or by extracting key parameters and result indicators from G-code execution logs using log parsing scripts. This transforms unstructured experience into computable model input.

[0027] Step S300: Based on the processing task dataset, perform task common-mode partitioning of the processing equipment topology network and obtain the common-mode partitioning results, wherein the common-mode partitioning results include multiple task common-mode partitioning classes.

[0028] The processing task dataset can be synonymous with the fine carving processing task dataset, representing the same object in different ways within the scheme, and can be used as input data for task commonality partitioning. Task commonality partitioning can be a clustering and classification process based on task feature similarity to identify task groups with common processing behavior patterns. The commonality partitioning result can be a set of classification labels output by the task commonality partitioning operation, identifying the commonality category to which each task belongs, and can be used as an index for constructing collaborative processing control strategies. A task commonality partitioning class can be a category unit in the commonality partitioning result, containing a group of tasks with similar processing characteristics, and can be used to define the application scope of a specific collaborative processing control strategy. In an exemplary embodiment, the task commonality partitioning class can include, but is not limited to, one or more of the following: high-hardness material low-roughness commonality class, three-dimensional freeform surface high-path-density commonality class, and microstructure high-positioning-accuracy commonality class.

[0029] Based on the processing task dataset, the task comorphism partitioning of the processing equipment topology network can be performed. This can be achieved by applying a clustering algorithm to the task dataset and classifying tasks according to feature similarity. Further, this operation can be implemented using K-means clustering with material properties, path complexity, and accuracy level as feature vectors, or by using hierarchical clustering combined with expert rules to constrain category boundaries, thereby identifying task groups with common processing behavior patterns. Obtaining the comorphism partitioning results can be achieved by reading the category labels and cluster center information output by the task comorphism partitioning operation. In a specific embodiment, this operation can be implemented by parsing a JSON-formatted category mapping table from the clustering model output file, or by directly obtaining the runtime clustering result object through memory sharing, thereby obtaining task classification criteria that can be used for policy matching.

[0030] Step S400: Construct a distributed collaborative processing controller by using the processing task dataset and the common-state partitioning results. The distributed collaborative processing controller includes multiple collaborative processing control strategies.

[0031] The distributed collaborative machining controller can be software or an embedded module deployed in the machining equipment topology network. It executes collaborative control logic based on common-mode partitioning and can dynamically generate and distribute control parameters adapted to the current task's common-mode class. In this embodiment, the distributed collaborative machining controller is jointly trained or configured from the machining task dataset and the common-mode partitioning results. Furthermore, the distributed collaborative machining controller can receive real-time machining status data as input, output control commands to each CNC device, and match corresponding strategies based on the task's common-mode partitioning class. The collaborative machining control strategy can be a set of device operating parameters and scheduling rules preset or learned for a specific task's common-mode partitioning class, which can be used to ensure consistent machining quality and efficiency for similar tasks across multiple devices. For example, the collaborative machining control strategy can include, but is not limited to, one or more of the following: tool feed-spindle speed coupling strategy, vibration suppression-coolant flow linkage strategy, path segmentation-device load balancing strategy, etc.

[0032] By combining the processing task dataset with the common-state partitioning results, a distributed collaborative processing controller can be constructed. This can be achieved by associating or training corresponding collaborative processing control strategies for each task's common-state partitioning class, and encapsulating these strategies into a callable controller module. In a specific embodiment, this operation can be implemented by configuring an if-then-style policy library based on a rule engine (e.g., enabling a low-feed, high-speed policy for high-hardness materials), or by using reinforcement learning to train a policy network on historical data and output continuous control parameters. This enables a mapping mechanism from task features to control strategies.

[0033] Step S500: Activate the processing equipment topology network to collect real-time processing status data of the target processing task area, input it into the distributed collaborative processing controller, and perform distributed collaborative control processing of multiple distributed CNC precision carving processing equipment.

[0034] The real-time machining status data can be multi-dimensional data reflecting the current operating status collected in real time by sensors of the CNC equipment during the machining process. This data can be used to drive the distributed collaborative machining controller to perform dynamic parameter adjustments and anomaly prediction. In this embodiment, the real-time machining status data is continuously sampled and uploaded through built-in sensors (such as current, vibration, temperature, and position encoders). For example, the real-time machining status data can include, but is not limited to, one or more of the following: spindle load current data, table vibration spectrum data, and online tool wear monitoring data. Distributed collaborative control machining of multiple distributed CNC engraving machines can be a machining process in which multiple machines synchronously execute parameter optimization and task allocation under the command of the distributed collaborative machining controller. This can be used to achieve cluster machining with load balancing, consistent quality, and optimal efficiency.

[0035] Activating the processing equipment topology network to collect real-time processing status data of the target processing task area can be achieved by starting the data acquisition modules of each device in the network and uploading status information at a preset frequency. Further, this operation can be implemented by subscribing to the real-time variable nodes of each device via OPCUA, or by deploying an edge agent program to periodically pull data from the device's PLC registers, thereby establishing a real-time feedback channel to support dynamic control. Inputting data into a distributed collaborative processing controller to perform distributed collaborative processing of multiple distributed CNC engraving machines can be achieved by inputting real-time processing status data into the controller, triggering strategy matching and parameter distribution, and driving the devices to operate collaboratively. In an exemplary embodiment, this operation can be achieved by the controller matching the common-state class of the current task, loading the corresponding strategy, adding real-time compensation, and then distributing it to each device; or by the controller dynamically evaluating the load distribution and reallocating sub-task path segments to balance device working hours, thereby realizing closed-loop collaborative processing based on common-state identification and real-time feedback.

[0036] Taking mass mixed production of metal casings for consumer electronics as an example, the distributed control processing method based on CNC precision carving tasks in this embodiment can be that in a certain 3C product production line, the target processing task area includes 8 CNC precision carving machines, which simultaneously undertake the processing tasks of mobile phone mid-frames made of aluminum alloy and stainless steel, with a precision requirement of ±5μm for both, but with significant differences in path complexity. The system first extracts 2,000 machining task data from the past three months from historical records to construct a precision carving task dataset. Through task commonality partitioning, it identifies two commonality partitioning categories: "high-gloss aluminum alloy planar milling" and "stainless steel deep cavity curved surface precision carving." Corresponding collaborative machining control strategies are configured for each category: the former adopts a high-speed, small-depth-of-cut strategy and limits the vibration threshold, while the latter enables a low-speed, high-rigidity strategy and enhances cooling. When a new order arrives, the scheduling platform assigns the task to the corresponding area, and the system automatically identifies the task characteristics and activates the corresponding strategy. Simultaneously, the topology network continuously collects spindle current and vibration data from each device. When the controller detects that the vibration of a certain device is close to the threshold, it immediately fine-tunes its feed rate and reallocates part of the path to idle devices, ultimately achieving efficient and stable mixed-line production of the two types of tasks without manual intervention to switch programs.

[0037] In one embodiment, reference Figure 2 An interactive machining task scheduling platform acquires multiple distributed CNC engraving machines within the target machining task area and constructs a machining day / night topology network, including: Step S101: Establish a communication connection with the processing task scheduling platform and obtain the target processing task configuration file; The target processing task configuration file can be a structured data file provided by the processing task scheduling platform, describing the target processing task area and its associated equipment information. It can serve as an authoritative data source for extracting equipment location, parameter, and status information. In an exemplary embodiment, the target processing task configuration file can be downloaded or subscribed to after establishing a communication connection with the scheduling platform, typically using JSON, XML, or a custom industrial format. Furthermore, the target processing task configuration file can include, but is not limited to, one or more of the following: static configuration files, dynamic task binding files, and equipment capability declaration files. Establishing a communication connection with the processing task scheduling platform and obtaining the target processing task configuration file can be achieved by initiating a connection request and receiving the configuration file data stream through a standard industrial communication protocol. Further, this operation can be implemented by using the HTTPSRESTAPI to call the scheduling platform interface to obtain the JSON format configuration file, or by subscribing to the task configuration nodes published by the scheduling platform through an OPCUA client, thereby achieving synchronization of task context and equipment resource information.

[0038] Step S102: Based on the target machining task configuration file, extract the location information, machining parameter information and current status information of multiple distributed CNC engraving machines within the target machining task area, and store them as basic information of the machining machines; The location information can be the coordinates or topological positioning identifiers of the distributed CNC engraving equipment in the physical space or logical production line, which can be used to determine the relative relationship of the equipment in the network and the feasibility of task path planning. For example, the location information may include, but is not limited to, one or more of the following: XYZ coordinates in the workshop coordinate system, production line station number, network IP address mapping location, etc. The machining parameter information can be a set of process capability parameters supported or currently set by the CNC equipment, such as spindle speed range, feed rate, stroke limit, etc., which can be used to reflect the processing capability boundaries of the equipment and support task matching and strategy generation. In a specific embodiment, the machining parameter information may include, but is not limited to, one or more of the following: kinematic parameters, dynamic parameters, accuracy level parameters, etc. The current status information can be the operating status of the equipment at the time of building the topology network, including idle, processing, fault, maintenance, etc., which can be used to determine whether the equipment can participate in the current task allocation. Further, the current status information may include, but is not limited to, one or more of the following: running status code, task queue length, health assessment indicators, etc. The basic information of the processing equipment can be a structured set of equipment metadata formed by integrating location information, processing parameter information, and current status information. This metadata can serve as a comprehensive representation of the equipment's capabilities and status, supporting subsequent topology construction and collaborative decision-making. In this embodiment, the basic information of the processing equipment can be extracted from the target processing task configuration file and stored in a standardized manner as a unified data structure. For example, the basic information of the processing equipment may include, but is not limited to, one or more of the following: static capability profiles, dynamic status snapshots, and mixed attribute records.

[0039] Based on the target machining task configuration file, the location information, machining parameter information, and current status information of multiple distributed CNC engraving machines within the target machining task area are extracted. This can be achieved by parsing fields in the configuration file and extracting three types of key information according to a preset schema. Furthermore, this operation can be implemented by using XPath or JSONPath to locate and extract fields from the configuration file, or by calling the built-in device information enumeration interface in the configuration file to read each item, thereby transforming unstructured or semi-structured configurations into structured attributes that can be used for modeling. The extracted three types of information are then stored as basic information for the machining machines. This can be achieved by integrating the three types of information into a unified data structure and persisting or caching it. Further, this operation can be implemented by writing to an in-memory device information hash table (using the device ID as the key), or by storing it in a lightweight embedded database (such as SQLite) device basic information table, thereby forming a standardized representation of device metadata.

[0040] Step S103: Analyze and determine the connection relationship and communication method of multiple distributed CNC engraving machines within the target processing task area, and construct a distributed processing equipment network; The connectivity relationships can be logical or physical communication or collaborative associations between multiple distributed CNC engraving machines, determining whether information can be transmitted between devices and affecting the accessibility of collaborative control. In a specific embodiment, connectivity relationships can include, but are not limited to, one or more of the following: wired Ethernet direct connection, wireless LAN adjacency, and indirect connection relayed through an edge gateway. The communication method can be the protocol or transmission mechanism used for data exchange between devices, affecting the real-time performance, reliability, and bandwidth of data transmission. For example, communication methods can use OPCUAoverTCP, ModbusTCP, MQTT message queues, etc. The distributed processing equipment network can be an initial logical network built based on connectivity relationships and communication methods, containing only device nodes and their interconnection structure, and can be used to provide the underlying connectivity skeleton for the processing equipment topology network. In this embodiment, the distributed processing equipment network can generate a graph structure by analyzing the physical wiring, network configuration, or process dependencies between devices. Furthermore, the distributed processing equipment network can be upgraded into a processing equipment topology network with attribute awareness capabilities after receiving basic processing equipment information injection.

[0041] Analyzing and determining the connection relationships and communication methods of multiple distributed CNC engraving machines within the target machining task area can be achieved by combining network scan results, configuration file declarations, or preset rules to infer the communication topology between devices. Furthermore, this operation can be implemented by performing LAN ARP scans and port probing to identify reachable devices and open protocols, or by directly constructing connection relationships based on the "communication group" field declared in the configuration file, thereby clarifying feasible paths and mechanisms for information flow. Constructing the distributed machining equipment network can be done by generating an initial network graph with devices as nodes and connection relationships as edges. Further, this operation can be achieved by using the NetworkX library to create an undirected graph (nodes being device IDs and edges being communication links), or by generating an adjacency matrix and storing it in shared memory for controller access, thereby establishing a skeleton model of the logical interconnection between devices.

[0042] Step S104: Update the basic information of the processing equipment to the distributed processing equipment network and obtain the processing equipment topology network.

[0043] The processing equipment topology network can be an enhanced equipment network model that integrates the structure of a distributed processing equipment network with basic information about the processing equipment. It can simultaneously express the interconnection relationships and individual capability states of equipment, supporting dynamic collaborative control. In this embodiment, the processing equipment topology network can be obtained by binding structured basic equipment information to corresponding nodes in the distributed processing equipment network. Furthermore, the processing equipment topology network can serve as the execution carrier for a distributed collaborative processing controller, providing channels for status reporting and command issuance. Updating the basic equipment information to the distributed processing equipment network can be achieved by attaching the basic equipment information as an attribute to the corresponding network node. Further, this operation can be implemented by traversing the list of basic equipment information and setting the attribute field for each network node, or by binding attributes to nodes using the MERGE statement in a graph database, thereby upgrading the network from a pure structural model to a state-aware model with attributes. Obtaining the processing equipment topology network can be achieved by outputting the distributed processing equipment network with completed attribute injection as the final topology network. Further, this operation can be implemented by returning a graph object reference for direct calling by subsequent modules, or by serializing it into a GraphML format file for offline analysis or backup, thereby obtaining a complete equipment network model that can be used for collaborative control.

[0044] For example, in a scenario of high-mix production of semiconductor packaging molds, the distributed control machining method based on CNC precision carving tasks in this embodiment can be as follows: In a precision mold workshop, a new order includes microstructure mold machining tasks made of three different materials (tungsten steel, ceramic, and alloy steel). The system first establishes an OPCUA connection with the scheduling platform to obtain a configuration file containing information on 8 CNC precision carving machines; it then extracts the location (e.g., areas A1 to A4), machining parameters (e.g., maximum speed 60krpm, minimum tool diameter 0.1mm), and current status (3 idle, 2 machining, 3 awaiting maintenance) of each machine; simultaneously, it confirms through network scanning that the machines in area A can be directly connected via gigabit Ethernet, while those in area B require a relay via an edge gateway; based on this, a distributed machining equipment network is constructed, and the above basic information is injected into each node to form a machining equipment topology network. This network is then used for subsequent common-mode partitioning, for example, assigning high-hardness tungsten steel tasks to machines A2 and A3, which have high-rigidity spindles and are idle, and enabling corresponding vibration suppression strategies, while ceramic tasks are routed to machine B1, which is equipped with a micro-diameter tool magazine. The entire process requires no manual intervention in equipment selection, achieving intelligent collaboration based on precise equipment profiling.

[0045] In one embodiment, based on the processing task dataset, a common-mode partitioning of the processing equipment topology network is performed, and the common-mode partitioning results are obtained, including: Preprocess the processing task dataset to obtain a standard processing task dataset.

[0046] The standard processing task dataset can be a structured, noise-free data set formed by cleaning, normalizing, and feature aligning the original processing task dataset. It can provide high-quality, computable input for task commonality partitioning algorithms. In this embodiment, the standard processing task dataset can be generated through preprocessing operations such as missing value imputation, outlier removal, unit unification, and feature encoding. For example, the standard processing task dataset can include, but is not limited to, one or more of the following: normalized numerical datasets, category label encoded datasets, and time-series path feature normalized datasets. Preprocessing the processing task dataset to obtain the standard processing task dataset can involve performing data cleaning, feature scaling, and format normalization processes. Furthermore, this operation can be achieved by using Z-score to normalize continuous features (such as spindle load and feed rate) and performing one-hot encoding on category features (such as material type), or by using a sliding window to smooth high-frequency noise in the path point sequence and extract statistical features (such as mean curvature and path length), thereby improving the accuracy and stability of subsequent clustering analysis.

[0047] Initialize the task common-mode partitioning algorithm and input the standard processing task dataset for common-mode partitioning to obtain the classification processing dataset, which includes multiple processing data groups.

[0048] The task commonality partitioning algorithm can be a computational model or rule system used to identify the inherent similarity of processing tasks and automatically cluster them. It can be used to divide a standard processing task dataset into subsets with common processing behavior patterns. In an exemplary embodiment, the task commonality partitioning algorithm can be implemented based on a hybrid method of unsupervised learning, semi-supervised clustering, or expert rule guidance. For example, the task commonality partitioning algorithm can employ K-means clustering, spectral clustering, autoencoder embedding clustering, etc. The categorized processing dataset can be a set of task data grouped by commonality category after being processed by the task commonality partitioning algorithm, which can be used as a direct data basis for constructing task commonality partitioning classes. In a specific embodiment, the categorized processing dataset can include, but is not limited to, one or more of the following: high material hardness-low surface roughness group, three-dimensional freeform surface-high path density group, microgroove array-high positioning repeatability group, etc. Multiple processing data groups can be each independent subset in the categorized processing dataset, containing several processing task records belonging to the same commonality category, which can be used to represent a historical experience set under a typical processing pattern. Furthermore, multiple processing data groups can be associated one-to-one with task commonality partitioning classes through relationships. Initializing the task commonality partitioning algorithm and inputting a standard processing task dataset for commonality partitioning can be achieved by loading clustering model parameters and using the standard dataset as input to perform the partitioning operation. Further, this operation can be implemented by setting the number of clusters K in K-means and determining the optimal value using the elbow rule, iterating until convergence, and then outputting cluster labels; or by training a variational autoencoder to map high-dimensional task features to a low-dimensional embedding space and then performing DBSCAN clustering in that space, thereby automatically discovering potential common patterns among tasks. Obtaining the classification processing dataset can be achieved by grouping and organizing the standard processing task dataset according to the category labels output by the commonality partitioning algorithm. Further, this operation can be implemented by writing data to different subdirectories by cluster ID (each directory corresponding to a processing data group), or by adding a class_id field to each record in the database and generating group views through SQLGROUPBY queries, thereby forming a clearly structured categorized historical experience database.

[0049] Based on the association between multiple processing data groups and multiple task common-state partitioning classes, multiple task common-state partitioning classes are mapped and established, and stored as common-state partitioning results.

[0050] The association relationship can be a mapping and binding logic between a processing data group and its corresponding task common-mode partitioning class, which can be used to assign semantic labels and control meanings to the data group. For example, the association relationship can include, but is not limited to, one or more of one-to-one static mapping, many-to-one generalized mapping, and dynamic threshold matching relationships. Multiple task common-mode partitioning classes can be task category units with technological semantics abstracted from the processing data groups through association relationships, and can be used as index keys for collaborative processing control strategies. In a specific embodiment, multiple task common-mode partitioning classes can include, but are not limited to, one or more of high-rigidity support requirement classes, micro-vibration sensitive classes, and high-speed path continuity classes. The common-mode partitioning result can be a final stored structured output containing all task common-mode partitioning classes and their corresponding processing data groups, which can be used by the distributed collaborative processing controller to achieve strategy matching. In this embodiment, the common-mode partitioning result can be obtained by serializing the category information established by mapping into a configuration file or database table. Based on the association relationship between multiple processing data groups and multiple task common-mode partitioning classes, mapping and establishing multiple task common-mode partitioning classes can be achieved by assigning a category name or identifier with technological semantics to each processing data group. Furthermore, this operation can be achieved by having process experts manually name each group of features (e.g., "stainless steel deep cavity precision carving class") and establish a manual mapping table, or by automatically parsing the mean of features within each group using a rule engine to generate descriptive labels (e.g., "high hardness + high curvature → rigidity priority class"). This transforms data-driven clustering results into interpretable and callable control semantic units. Storage as common-state partitioning results can involve persistently saving the task common-state partitioning classes and their associated processing data groups. Further, this operation can be achieved by serializing the data into a JSON file (containing category IDs, semantic labels, feature center vectors, and a list of member tasks) or writing it into a graph database (with "common-state classes" as nodes and "containing tasks" as edges), thus forming reusable and updatable knowledge assets.

[0051] Taking the mixed production of precision medical device parts using multiple materials as an example, the distributed control machining method based on CNC precision carving tasks in this embodiment can be as follows: A medical parts factory needs to simultaneously process titanium alloy fasteners, PEEK polymer connectors, and stainless steel pipe clamps. The material properties, path complexity, and precision requirements of these three components differ significantly. The system first preprocesses 1,500 machining records from the past six months: abnormal tasks such as tool breakage are removed, and the measured values ​​of material hardness, path points, and surface roughness are normalized using Min-Max to form a standard machining task dataset. Subsequently, a spectral clustering algorithm is initialized, and the tasks are divided into four machining data groups based on Laplacian feature mapping. After analysis, one group is characterized by "high elastic modulus + low thermal conductivity + high precision" and is mapped and named "Titanium Alloy Microstructure Common State Class"; another group is characterized by "low rigidity + high thermal sensitivity + medium curvature" and is named "PEEK Lightweight Structure Common State Class". These categories, along with their feature centers and historical task lists, are stored as the common state partitioning results. When a new order arrives, the system extracts its task characteristics, matches it to the nearest common state class, and automatically loads the corresponding collaborative control strategy. For example, it enables a combination of low-temperature cooling and low-vibration feed for titanium alloys, and limits the accumulation of cutting heat and enhances support rigidity for PEEKs. The entire process does not require engineers to rewrite G-code or adjust parameter templates.

[0052] In one embodiment, before performing task common-mode partitioning of the processing equipment topology network based on the processing task dataset, the method further includes: Analyze the processing task dataset to obtain the key feature set for task commonality partitioning; The key feature set is divided according to task commonality, and the dimensionality reduction and size reduction processing of the processing task dataset is carried out.

[0053] The key feature set for task commonality partitioning can be a subset of core features selected from the processing task dataset that has a decisive influence on task commonality partitioning. It can be used as the input dimension for task commonality partitioning, improving the physical meaning and engineering effectiveness of the clustering results. In an exemplary embodiment, the key feature set for task commonality partitioning can identify variables strongly correlated with processing quality, efficiency, and stability through feature importance assessment, correlation analysis, or domain knowledge guidance. Furthermore, the key feature set for task commonality partitioning can include, but is not limited to, one or more of the following: material-process coupling feature set, geometric path topology feature set, and precision-surface quality constraint feature set. Dimensionality reduction processing can be a data preprocessing operation based on the key feature set for task commonality partitioning, performing dimensionality compression and redundancy removal on the original processing task dataset. This can reduce the computational complexity of subsequent commonality partitioning, improve clustering accuracy, and avoid model distortion caused by high-dimensional sparsity. For example, dimensionality reduction processing can retain key features, remove irrelevant or weakly correlated variables, and further merge relevant dimensions through linear or nonlinear transformations. In one specific embodiment, dimensionality reduction processing can employ principal component analysis (PCA) to project high-dimensional features into a low-dimensional orthogonal space, or use feature selection methods based on information gain or L1 regularization to remove redundant fields.

[0054] Analyzing the processing task dataset to obtain a key feature set for task commonality partitioning can be achieved by performing statistical analysis or machine learning evaluation on the feature variables in the dataset to screen out key features with high discriminative power for task behavior patterns. Further, this operation can be achieved by using random forest or XGBoost models to output feature importance rankings and selecting the top N highly important features, or by defining the range of key features using expert rules (e.g., retaining only material type, path point density, tolerance level, etc.), thereby focusing on the essential factors affecting processing commonality and eliminating noise and redundant interference. Using the key feature set for task commonality partitioning to perform dimensionality reduction processing on the processing task dataset can be done by projecting or mapping the original dataset to a low-dimensional subspace defined by the key feature set, generating a simplified task data representation. Further, this operation can be achieved by directly pruning the original data table to retain only the key feature columns to form a new dataset, or by using an autoencoder network to learn the low-dimensional embedding representation of key features and reconstruct the task data, thus reducing data size and dimensionality while preserving task differences, improving the efficiency and robustness of subsequent commonality partitioning.

[0055] Taking a multi-material mixed production scenario for semiconductor packaging molds as an example, the distributed control machining method based on CNC precision carving tasks in this embodiment can be as follows: A precision mold workshop has a historical record containing 5,000 machining tasks, each record containing more than 30 original fields (such as ambient temperature and humidity, operator ID, equipment model, tool batch, etc.). Before performing task commonality partitioning, the system first analyzes the dataset. Through feature importance assessment, it finds that only four items—"material hardness," "mean surface curvature," "minimum feature size," and "surface roughness target"—contribute more than 85% to the final machining stability, constituting the key feature set for task commonality partitioning. Subsequently, the original dataset is subjected to dimensionality reduction processing using this key feature set, removing the remaining 26 low-relevance fields, and standardizing continuous variables. The data after this processing is input into a clustering algorithm, which clearly separates three types of commonality tasks: "superhard alloy microstructures," "soft metal high-gloss surfaces," and "composite material transition zones." Their contour coefficients are significantly better than the results without dimensionality reduction. The subsequent controller, based on this category's rapid matching strategy, eliminates the need for manual adjustments when switching between different packaging mold orders, and improves the accuracy of vibration anomaly warnings, supporting stable production with high flexibility and low scrap.

[0056] In one embodiment, the processing task dataset is analyzed to obtain a set of key features for task commonality partitioning, including: Based on the principle of standardization, the dataset for processing tasks is decentralized.

[0057] This operation can be performed by subtracting the mean from each processing indicator column in the processing task dataset, making the mean of each dimension zero. Furthermore, this operation can be achieved by performing z-score decentering on continuous indicators (subtracting only the mean, not the standard deviation) or by keeping the original values ​​of categorical coded indicators and decentering only numerical fields. This eliminates the differences in dimensions and biases between different indicators, satisfying the prerequisites for principal component analysis.

[0058] Based on the decentralized results, a standard processing data matrix is ​​constructed, with the number of processing tasks in the processing task dataset as the number of rows in the matrix and the number of processing indicator categories in the processing task dataset as the number of columns in the matrix.

[0059] The standard machining data matrix can be a numerical matrix organized by the number of tasks in the decentralized machining task dataset, with rows representing the number of tasks and columns representing the number of machining index categories. This matrix can provide structured input for subsequent correlation analysis and principal component extraction. In this embodiment, the standard machining data matrix can represent each machining task as a row and each machining index as a column, forming an m×n real number matrix. For example, the standard machining data matrix can employ sparse matrix storage to handle high-dimensional, low-fill-rate scenarios, or it can be divided into sub-matrices by time windows to support incremental processing. Furthermore, the standard machining data matrix can include, but is not limited to, one or more of the following: material-process coupling matrix, geometry-precision joint matrix, and dynamic cutting parameter matrix.

[0060] Establish the correlation matrix of the standard processing data matrix, calculate the eigenvalues ​​and eigenvectors of the correlation matrix, and obtain the feature set of the correlation matrix.

[0061] The correlation matrix can be a symmetric square matrix composed of Pearson correlation coefficients among the processing indicators calculated from the standard processing data matrix. It can be used to characterize the linear dependencies between different processing indicators and for principal component orientation identification. In an exemplary embodiment, the correlation matrix can be obtained by calculating and normalizing the covariances between the column vectors of the standard processing data matrix. Furthermore, the correlation matrix can be calculated using the Pearson correlation coefficient to obtain a fully connected correlation matrix, or by employing regularized correlation estimation (such as Ledoit-Wolf) to improve small-sample stability, thereby revealing the linear association structure among processing indicators and providing a basis for principal component orientation identification.

[0062] The correlation matrix feature set can be a set of eigenvalues ​​and corresponding eigenvectors obtained after eigenvalue decomposition of the correlation matrix. It can be used to construct the basis of the principal component space, where each eigenvector represents a potential common-mode influence direction. In a specific embodiment, the correlation matrix feature set can be obtained by performing spectral decomposition on the correlation matrix to solve for all its eigenvalues ​​(eigenvalues) and corresponding unit eigenvectors. Furthermore, the correlation matrix feature set can be obtained by using the QR iterative method to solve for the complete feature system, or by using the power method or Lanczos algorithm to calculate only the first k largest eigenvalue pairs to accelerate the implementation, thereby obtaining the variance magnitude and direction information of the principal components. For example, the correlation matrix feature set can include, but is not limited to, one or more of the following: high-variance principal component sets, medium-explanatory-power component sets, and noise-dominated component sets.

[0063] Based on the variance contribution scores of the feature set of the relevant matrix, a cumulative variance contribution sequence is obtained by accumulating the variance contribution scores. The cumulative variance contribution sequence includes multiple cumulative variance contribution scores, and each cumulative variance contribution score has a processing feature class label.

[0064] The variance contribution cumulative sequence can be a monotonically increasing sequence formed by sequentially accumulating the variance contribution scores of each principal component after eigenvalues ​​are sorted in descending order. This sequence can be used to quantify the proportion of total information covered when retaining the top k principal components. In this embodiment, the variance contribution cumulative sequence can be obtained by normalizing the eigenvalues ​​to variance contribution scores and then performing prefix accumulation. Furthermore, the variance contribution cumulative sequence can be implemented by dynamically generating a cumulative curve and visualizing it for manual verification, or by automatically recording the principal component index corresponding to each cumulative score. This provides a quantifiable information retention assessment tool and supports dimensionality truncation decisions. For example, the variance contribution cumulative sequence can include, but is not limited to, one or more of the following: fast-converging cumulative sequences, slowly changing cumulative sequences, and cumulative sequences with a clear plateau period.

[0065] Based on the preset contribution rate constraint, the cumulative variance contribution score is filtered according to the cumulative variance contribution sequence, and the processing feature class label of the filtering result is extracted to generate the task commonality partitioning key feature set.

[0066] In this operation, based on a preset contribution rate constraint, the cumulative variance contribution score is filtered according to the cumulative variance contribution sequence. This can be achieved by setting a threshold (e.g., 0.85) and selecting the minimum k such that the cumulative score of the first k terms is greater than or equal to the threshold. Furthermore, this filtering operation can be implemented by using a fixed threshold (e.g., 90%) for hard truncation, or by combining inflection point detection (elbow method) to adaptively determine the k value, thus achieving a balance between information preservation and dimensionality compression. The processing feature class labels of the filtering results are extracted to generate a key feature set for task commonality partitioning. This can be achieved by analyzing the feature vector loadings of the selected principal components and extracting the original processing index class labels with larger absolute loading values. Further, this extraction operation can be implemented by setting a loading threshold (e.g., |loading| > 0.5) to filter significant features, or by taking the union of the significant features of multiple principal components as the final key feature set, thus mapping abstract principal components back to interpretable engineering features, forming key inputs that can be used for commonality partitioning.

[0067] The task commonality partitioning key feature set can be a set of original processing feature class labels corresponding to principal components selected through the variance contribution accumulation sequence under a preset contribution rate constraint. This set can be used as low-dimensional, high signal-to-noise ratio input features for task commonality partitioning, improving the physical meaning and engineering effectiveness of clustering. In a specific embodiment, the task commonality partitioning key feature set can be obtained by backtracking the original variables with larger loadings among the selected principal components and extracting their corresponding processing feature class labels. Furthermore, the task commonality partitioning key feature set can rely on the variance contribution accumulation sequence and the preset contribution rate constraint to jointly determine the screening boundary; its content is determined by the loading structure of high-contribution principal components in the correlation matrix feature set.

[0068] Taking the production scenario of high-precision molds with mixed indicators as an example, the distributed control machining method based on CNC precision carving tasks in this embodiment can be: a mold factory's historical data contains 1200 precision carving tasks, each containing 20 indicators (such as material hardness HRC, feed speed mm / min, mean path curvature, surface roughness Ra, spindle power W, etc.). The system first decentralizes the data, constructing a 1200×20 standard machining data matrix; it then calculates its 20×20 correlation matrix, finding that material hardness is highly correlated with spindle power (r=0.89), while path curvature is negatively correlated with feed rate (r=-0.76); through feature decomposition, the cumulative contribution rate of the five principal components reaches 91.3%, with the first principal component showing high loads on material hardness, spindle power, and tool wear rate, and the second principal component showing significant loads on path curvature, feed rate, and vibration amplitude; setting a contribution rate constraint of 90%, the system selects the top five principal components and extracts the original index class labels whose absolute load values ​​are greater than 0.6, ultimately generating a task commonality partitioning key feature set containing six items: "material hardness," "spindle power," "path curvature," "feed rate," "vibration amplitude," and "tool wear rate." This feature set eliminates irrelevant variables such as ambient temperature and operator number, enabling subsequent common-state partitioning to accurately identify real process modes such as "high-hardness low-speed steady-state class" and "high-bend speed-changing dynamic class", supporting the controller's precise matching strategy and avoiding erroneous clustering caused by redundant feature interference.

[0069] In one embodiment, a distributed collaborative processing controller is constructed using the processing task dataset and the common-state partitioning results, including: Randomly select from the comorphic partitioning results to obtain the first task comorphic partitioning class.

[0070] The first task common-state partitioning class can be a specific task common-state category randomly selected from the common-state partitioning results. It represents a subset of tasks with similar processing characteristics and can be used as an anchor point for the target task type in constructing a specific collaborative control strategy. In an exemplary embodiment, the first task common-state partitioning class can include, but is not limited to, one or more of the following: high thermal conductivity ceramic microstructure fine carving, low-stress milling of titanium alloy thin-walled parts, and ultra-smooth polishing of glass substrates. Randomly selecting the first task common-state partitioning class from the common-state partitioning results can be achieved by selecting a category from multiple unprocessed task common-state partitioning classes according to a random strategy. Furthermore, this operation can be implemented by using a pseudo-random number generator to select the category index according to a uniform distribution, or by combining a weighted average of the category sample sizes with probability sampling to prioritize coverage of high-frequency task types, thereby initiating the strategy construction process for a specific common-state category.

[0071] Based on the first task common-state classification, and based on the common-state classification result, the first processing data group is invoked.

[0072] The first machining data set can be a subset of historical machining tasks corresponding to the common-state partitioning class of the first task. It includes the input features and successful machining result labels of that task class, and can be used to provide training samples highly consistent with the target task scenario for supervised training. In a specific embodiment, the first machining data set may include, but is not limited to, one or more of the following: a data set containing spindle load-surface roughness mapping relationships, a data set containing path density-feed speed-vibration amplitude triplets, and a data set containing material hardness-tool wear rate-cooling intensity correlation records. For example, the first machining data set can be composed of matching samples selected from the engraving machining task dataset based on the category labels in the common-state partitioning results.

[0073] Based on the common-state classification of the first task, and the results of the common-state classification, the first processing data group is retrieved. This can be done by retrieving all task records belonging to the first task common-state classification class from the fine carving task dataset, based on the identifier of that class. Furthermore, this operation can be achieved by executing an SQL query to filter records with label='first task common-state classification class' from the database, or by traversing the task data list in memory and filtering by category label to form a subset. This allows obtaining a subset of training data that strictly matches the target common-state class.

[0074] Based on the first set of processed data, a first collaborative processing control strategy is obtained through supervised training.

[0075] The supervised training method can be a machine learning approach that uses labeled historical processing data to learn the mapping relationship between input features and optimal control output, and can be used to generate intelligent control strategies that can be generalized to similar tasks. In an exemplary embodiment, the supervised training method can employ decision tree-based rule extraction, neural network-based end-to-end mapping learning, or Gaussian process regression-based uncertainty modeling. Furthermore, the supervised training method can use process parameters and equipment status from the first processing data set as input, and actual, successfully adopted control commands or parameter configurations as output labels for model training.

[0076] Based on the first set of processing data, a first collaborative processing control strategy is obtained through supervised training. This can be achieved by using the first set of processing data as a training set and employing a supervised learning algorithm to fit a mapping function from input (such as material properties and path complexity) to output (such as equipment parameters and scheduling scheme). Furthermore, this operation can be implemented by training a multilayer perceptron model (with standardized task feature vectors as input and continuous control parameters such as spindle speed and feed rate as output) or by training a classification-regression hybrid model (first determining whether vibration suppression mode needs to be enabled, and then outputting the corresponding parameter combination). This allows for the generation of intelligent control strategies reusable for this type of common-state task.

[0077] The first collaborative processing control strategy is stored in the distributed collaborative processing controller.

[0078] The first collaborative processing control strategy can be a dedicated control logic or parameter generation model obtained through supervised training for the first task commonality partitioning class. It can be used to provide customized, data-driven collaborative control instructions when tasks of this commonality partitioning class are executed. Furthermore, after the first collaborative processing control strategy is stored in the distributed collaborative processing controller, it can be invoked and executed when similar tasks are identified. Storing the first collaborative processing control strategy in the distributed collaborative processing controller can be achieved by persistently saving the trained strategy model or rule set to the strategy library of the distributed collaborative processing controller and establishing a mapping index with the first task commonality partitioning class. Further, this operation can be implemented by serializing the model into an ONNX format file and storing it in the controller's local storage and updating the strategy registry, or by compiling the rule logic into an executable script module and dynamically loading it into the controller's runtime environment, thereby enabling the strategy to be automatically invoked during subsequent scheduling of similar tasks.

[0079] Taking the micro-hole array processing of semiconductor packaging substrates as an example, the distributed control processing method based on CNC precision carving tasks in this embodiment can be as follows: In an advanced packaging production line, the system has completed the common-state classification of 5000 historical micro-hole processing tasks, identifying six common-state categories, such as "FR-4 substrate high-density micro-hole type" and "ceramic substrate depth-to-diameter ratio > 10 type". During the controller construction stage, the system randomly selects "FR-4 substrate high-density micro-hole type" as the first task common-state classification category, and retrieves 820 successful processing records to form the first processing data group. This data group includes fields such as drilling depth, hole spacing, spindle current, cooling gas pressure, and final hole wall quality rating. A gradient boosting tree model is used for supervised training to learn how to dynamically adjust the feed rate and chip removal frequency to avoid burrs under different hole densities. After training, the strategy is named "Strategy_FR4_HighDensity_V1" and stored in the distributed collaborative processing controller. When a new order arrives and is identified as a similar task, the controller automatically activates the strategy, fine-tuning it by integrating current equipment vibration data in real time to ensure that the consistency of thousands of micro-holes reaches ±2μm, while controlling the tool life prediction error to within 5%.

[0080] In one embodiment, distributed collaborative control processing of multiple distributed CNC precision carving machines includes: Analyze real-time processing status data to determine the common-state classification of real-time tasks.

[0081] The real-time task common-state classification can be a common-state category dynamically determined based on real-time processing status data, used for online matching of historically constructed collaborative control strategies. In this embodiment, the real-time task common-state classification can be obtained by inputting real-time processing status data into a pre-trained classification model or rule engine and comparing its similarity with the feature centers of historical task common-state classifications. Furthermore, the real-time task common-state classification can serve as an index for control matching, driving the distributed collaborative processing controller to select the corresponding collaborative processing control strategy. For example, the real-time task common-state classification can include, but is not limited to, one or more of the following: real-time common-state classification based on spindle load characteristics, real-time common-state classification based on vibration spectrum patterns, and real-time common-state classification based on path execution deviation.

[0082] Analyzing real-time processing status data to determine the common-state classification of real-time tasks can involve feature extraction and classification reasoning of the collected real-time processing status data, outputting the common-state category label to which the current task belongs. In an exemplary embodiment, this operation can be achieved by using a lightweight neural network model for real-time reasoning at edge nodes; in a specific embodiment, it can also be achieved by logical judgment based on a preset feature threshold range and rule tree, thereby enabling the system to extend from static historical common-state classification to dynamic online identification, improving the system's perception accuracy of the running task status.

[0083] Based on the real-time task common-state classification, multiple collaborative processing control strategies in the distributed collaborative processing controller are traversed for control matching.

[0084] The multiple collaborative machining control strategies can be sets of control logic stored internally by the distributed collaborative machining controller, each corresponding to a different common-state classification of the task. These sets provide differentiated and optimized operating parameters and scheduling rules for different common-state tasks. For example, the multiple collaborative machining control strategies may include, but are not limited to, tool wear compensation control strategies, thermal deformation suppression control strategies, and multi-axis synchronization optimization control strategies.

[0085] Based on the real-time task common-state classification, multiple collaborative processing control strategies in the distributed collaborative processing controller are traversed for control matching. This can be achieved by using the real-time task common-state classification as the query key to retrieve or evaluate the most suitable collaborative processing control strategy in the strategy library. Furthermore, this operation can be implemented by directly indexing pre-associated strategies through hash mapping; in a specific embodiment, similarity calculation can also be used to sort all strategies and select the Top-K strategies, thereby ensuring that the control logic highly matches the actual state of the current task and avoiding performance degradation caused by strategy mismatch.

[0086] Based on the control matching results, the distributed collaborative processing controller is activated to make processing control decisions.

[0087] The control matching result can be the optimal strategy identifier or parameter set output after comparing the real-time task common-state classification with multiple collaborative processing control strategies, serving as a direct input basis for activating processing control decisions. In this embodiment, the control matching result can determine the best match through nearest neighbor matching, confidence threshold judgment, or a weighted scoring mechanism. For example, the control matching result can include, but is not limited to, exact matching results, fuzzy approximate matching results, and multi-strategy fusion matching results.

[0088] Machining control decisions can be a set of specific equipment operation instructions generated by a distributed collaborative machining controller based on control matching results, used to guide multiple CNC machines to execute consistent and optimized machining actions. In an exemplary embodiment, machining control decisions are output after fine-tuning by invoking the matched collaborative machining control strategy and combining it with the current real-time status data. Furthermore, activating the distributed collaborative machining controller for machining control decisions based on the control matching results can be achieved by invoking the parameter template corresponding to the strategy and superimposing real-time compensation quantities; in a specific embodiment, it can also be achieved by triggering the state machine process built into the strategy, outputting step-by-step control instructions in stages, thereby completing the transformation from strategy matching to actual control output and forming executable control actions.

[0089] The processing equipment topology network distributes the processing control decision results to multiple distributed CNC precision carving processing equipment for distributed collaborative control processing.

[0090] The machining equipment topology network distributes machining control decisions to multiple distributed CNC engraving machines for distributed collaborative control machining. This can be achieved by broadcasting or directing the machining control decisions to the target machines via the communication links of the topology network. In one specific embodiment, this operation can be implemented by pushing unified baseline parameters to all devices using a publish-subscribe model; alternatively, it can be implemented by sending differentiated instructions only to the devices undertaking sub-tasks based on task allocation relationships, thereby achieving cluster-level synchronous execution and ensuring consistency and collaboration in multi-machine machining.

[0091] Taking the precision engraving production of semiconductor packaging molds using mixed materials as an example, the distributed control machining method based on CNC precision engraving tasks in this embodiment can be as follows: A factory simultaneously processes packaging molds made of tungsten steel and ceramic materials. The paths for both are similar, but the material removal characteristics differ greatly. During the processing, the system continuously collects spindle current, vibration, and temperature data. When a sudden increase in spindle load and a shift in vibration spectrum are detected, the analysis module classifies it into the "highly brittle material microcrack sensitive common state class." The controller then traverses the strategy library and matches the "low-impact feed + high-frequency cooling pulse" strategy designed specifically for this type of material. A machining control decision is generated, which includes a 15% reduction in feed rate and an increase in coolant pressure. This decision is then sent to the current equipment and adjacent standby equipment through the machining equipment topology network, allowing the latter to preheat the spindle in preparation for subsequent machining. The entire process requires no machine downtime or manual intervention, effectively avoiding chipping defects in ceramic parts while maintaining stable production line cycle time.

[0092] In one embodiment, the distributed collaborative control process of multiple distributed CNC engraving machines further includes: The system periodically collects data on actual machining dimensions and tool wear from various distributed CNC engraving equipment, and then compiles this data to generate a real-time machining error dataset.

[0093] The actual machining dimensions can be measured values ​​of key geometric dimensions of the workpiece obtained through online measurement or feedback systems after machining by CNC engraving equipment. These dimensions can serve as a direct basis for evaluating machining accuracy deviations. In this embodiment, the actual machining dimensions can be acquired through a laser probe, contact probe, or vision measurement module integrated into the equipment and compared with the theoretical CAD model. For example, the actual machining dimensions can include, but are not limited to, one or more of the following: measured contour dimensions, measured hole coordinates, and measured surface height data. Tool wear data can be a quantitative indicator reflecting the degree of edge wear during current tool use. This data can be used to predict machining quality deterioration trends and participate in error attribution analysis. Furthermore, tool wear data can be indirectly estimated through spindle current change rate, cutting force model inversion, or tool life counter, or it can be directly measured by a dedicated detection device. In an exemplary embodiment, tool wear data can include, but are not limited to, one or more of the following: radial wear, flank wear band width, and edge dulling index.

[0094] A real-time machining error dataset can be a structured data set representing the machining deviation of the current batch, generated by fusing periodically collected actual machining dimensions and tool wear data. It can be used to support statistical analysis and common-state stability assessment of machining errors. In one specific embodiment, the real-time machining error dataset can be obtained by aligning and standardizing the dimensional deviation and wear data fed back from each device according to task ID, timestamp, and device ID. For example, the real-time machining error dataset can include, but is not limited to, one or more of the following: dimension deviation-dominated error set, wear-related error set, and multi-dimensional coupled error set. Periodically collecting actual machining dimensions and tool wear data fed back from each distributed CNC engraving machine can be done at fixed time intervals or based on the number of workpieces processed, obtaining dimensional measurement results and tool status data from the device. Furthermore, this operation can be achieved by triggering online measurement and uploading data after each workpiece is completed, or by collecting a batch of data and summarizing it after processing N workpieces, thereby establishing a closed-loop feedback channel for machining quality and achieving continuous monitoring of actual machining deviations. The real-time machining error dataset can be generated by aggregating dimensional deviation and wear data from multiple devices according to common-state categories to form structured error records. Furthermore, this operation can be implemented by storing data in the form of a database table (fields include device ID, task ID, theoretical value, measured value, wear amount, and timestamp) or by maintaining it in real time in a DataFrame structure in memory (supporting streaming computing), thereby providing a data foundation for error statistics and common-state stability assessment.

[0095] Calculate the average overall machining error under the current real-time task common-state partitioning. When the average overall machining error exceeds the preset machining accuracy threshold, re-partition the current machining state data into common-state data, re-match the collaborative machining control strategy, and update and issue machining control commands to complete the dynamic correction of machining control.

[0096] The overall average processing error can be the statistical average of the processing errors of all participating devices under the current real-time task common-state classification, and can be used as a quantitative criterion for determining whether the current common-state control strategy has failed. In this embodiment, the overall average processing error can be obtained by calculating the arithmetic mean or weighted average of the task error items belonging to the same common-state class in the real-time processing error dataset. For example, the overall average processing error can include, but is not limited to, one or more of the following: the average absolute error, the average relative tolerance percentage, and the average directional deviation. The preset processing accuracy threshold can be the upper limit of the maximum allowable average processing error set according to process requirements, and can be used as the judgment boundary for triggering the dynamic correction mechanism. In a specific embodiment, the preset processing accuracy threshold can be configured by the process engineer based on the product tolerance zone, historical yield data, or customer standards. For example, the preset processing accuracy threshold can include, but is not limited to, one or more of the following: a static fixed threshold, a material adaptive threshold, and a batch size adjustment threshold.

[0097] The current machining status data can be multi-dimensional operational status data collected at the moment the error exceeds the limit, used for re-common state partitioning. It can be used as input for a new round of common state identification, reflecting the current actual machining conditions. For example, the current machining status data can include, but is not limited to, one or more of the following: wear-compensated status data, thermal deformation-corrected status data, and path execution offset status data. The new common state partitioning result can be the classification output obtained by re-performing the task common state partitioning based on the updated current machining status data after the error exceeds the limit. It can be used to replace the original common state partitioning result to match the drifted machining state. For example, the new common state partitioning result can include, but is not limited to, one or more of the following: wear-sensitive new common state class, thermal drift-dominated new common state class, and material batch variation-type new common state class. The updated collaborative machining control strategy can be a control strategy re-matched or generated based on the new common state partitioning result, used to replace the original strategy. It can be used to adapt to the current actual machining conditions and restore machining accuracy stability. For example, the updated collaborative machining control strategy can include, but is not limited to, one or more of the following: tool compensation enhancement strategy, thermal stability priority strategy, and low-stress cutting strategy.

[0098] Machining control commands can be specific parameter adjustment or action commands generated and issued to the CNC equipment by a distributed collaborative machining controller, and can be used to execute dynamically corrected control logic. For example, machining control commands can include, but are not limited to, one or more of the following: spindle speed adjustment commands, feed rate correction commands, and cooling strategy switching commands. Calculating the overall machining error mean under the current real-time task common-state classification can be achieved by filtering all samples belonging to the current common-state class in the real-time machining error dataset and calculating their average error. Further, this operation can be implemented by using a simple arithmetic mean or a weighted average based on equipment weight or workpiece importance, thereby quantifying the overall machining performance under the current common-state control strategy. Determining whether the overall machining error mean exceeds a preset machining accuracy threshold can be achieved by comparing the calculated overall machining error mean with the preset threshold. Further, this operation can be triggered by a single exceedance triggering reclassification, or by two consecutive exceedances or the exceedance ratio within a sliding window reaching the threshold, thereby determining whether to initiate the dynamic correction process.

[0099] Re-partitioning the current machining state data into common states can be achieved by using the updated current machining state data as input and performing the same clustering or classification process as the initial common state partitioning. Furthermore, this operation can be implemented through incremental clustering (fine-tuning based on the original common state centers) or full re-clustering (ignoring historical partitions and reconstructing the common state structure entirely based on current data), thereby identifying state drift caused by tool wear, temperature drift, or material variation, and generating common state categories adapted to the new working conditions. Re-matching the collaborative machining control strategy can be done by searching for or generating a corresponding collaborative machining control strategy in the strategy library based on the new common state partitioning results. Furthermore, this operation can be achieved by directly mapping to a pre-stored strategy or by fine-tuning parameters based on similar historical strategies to generate a new strategy, thereby ensuring that the control logic is consistent with the current actual machining state. Updating and issuing machining control commands to complete the dynamic correction of machining control can be achieved by converting the updated collaborative machining control strategy into specific control commands and issuing them to relevant equipment through the topology network. Furthermore, this operation can be achieved by immediately overwriting the original instructions and synchronously restarting the machining cycle, or by a smooth transition (gradually adjusting parameters to the target value to avoid shock), thereby enabling online optimization of machining parameters and restoration of accuracy.

[0100] Taking the continuous production of high-precision optical molds as an example, the distributed control machining method based on CNC engraving tasks in this embodiment can be as follows: A production line uses 6 CNC engraving machines to continuously process glass mold cores, requiring a surface shape accuracy of ≤2μm. Every 10 pieces processed, the system collects the online measured dimensions of each machine and the tool wear amount calculated by the spindle load, generating a real-time machining error dataset. In the initial stage, the task is classified into the "ultra-precision milling common state class" and a high-rigidity, low-vibration strategy is enabled. When running to the 50th piece, the overall average machining error rises to 2.3μm, exceeding the preset threshold of 2.0μm; the system then extracts the current machining state data containing wear aggravation characteristics, re-clusters and finds that the task has drifted to the "tool passivation sensitive common state class"; the controller matches the preset "wear compensation enhancement strategy", automatically reduces the feed rate by 8%, increases the cooling pressure, and activates tool path micro-compensation; after the new instruction is issued, the error of subsequent workpieces falls back to 1.7μm, without the need for machine stoppage or manual intervention, ensuring the accuracy stability of mass production.

[0101] Furthermore, to achieve the above objectives, the present invention also provides a distributed control machining device based on CNC precision carving tasks. The device includes: a memory, a processor, and a distributed control machining program based on CNC precision carving tasks stored in the memory and executable on the processor. The distributed control machining program based on CNC precision carving tasks is configured to implement the steps of the distributed control machining method based on CNC precision carving tasks as described above.

[0102] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium storing a distributed control machining program based on a CNC engraving task. When the distributed control machining program based on the CNC engraving task is executed by a processor, it implements the steps of the distributed control machining method based on a CNC engraving task as described above.

[0103] Other embodiments or specific implementations of the distributed control machining equipment based on CNC precision carving tasks described in this invention can be referred to the above-described method embodiments, and will not be repeated here.

[0104] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A distributed control machining method based on CNC precision carving tasks, characterized in that, The method includes: An interactive machining task scheduling platform acquires multiple distributed CNC precision carving machines in the target machining task area and constructs a machining equipment topology network. Based on the historical fine carving records of the target processing task area, extract the fine carving task dataset; Based on the processing task dataset, the processing equipment topology network is partitioned into task common modes, and the common mode partitioning results are obtained, wherein the common mode partitioning results include multiple task common mode partitioning classes; A distributed collaborative processing controller is constructed using the processing task dataset and the comorphic partitioning results, wherein the distributed collaborative processing controller includes multiple collaborative processing control strategies; The processing equipment topology network is activated to collect real-time processing status data of the target processing task area, which is then input into the distributed collaborative processing controller to perform distributed collaborative control processing of multiple distributed CNC precision carving processing equipment.

2. The distributed control machining method based on CNC precision carving tasks as described in claim 1, characterized in that, The interactive machining task scheduling platform acquires multiple distributed CNC engraving machines in the target machining task area and constructs a machining machine topology network, including: Establish a communication connection with the processing task scheduling platform to obtain the target processing task configuration file; Based on the target machining task configuration file, the location information, machining parameter information and current status information of multiple distributed CNC precision carving machines within the target machining task area are extracted and stored as basic information of the machining machines. Analyze and determine the connection relationships and communication methods of multiple distributed CNC precision carving machines within the target machining task area, and construct a distributed machining equipment network; The basic information of the processing equipment is updated to the distributed processing equipment network to obtain the processing equipment topology network.

3. The distributed control machining method based on CNC precision carving tasks as described in claim 2, characterized in that, The step of performing task comorphic partitioning of the processing equipment topology network based on the processing task dataset and obtaining the comorphic partitioning results includes: The processing task dataset is preprocessed to obtain a standard processing task dataset; Initialize the task common-mode partitioning algorithm and input the standard processing task dataset for common-mode partitioning to obtain a classified processing dataset, wherein the classified processing dataset includes multiple processing data groups; Based on the association between multiple processing data groups and multiple task comorphic partitioning classes, multiple task comorphic partitioning classes are mapped and established, and stored as the comorphic partitioning results.

4. The distributed control machining method based on CNC precision carving tasks as described in claim 1, characterized in that, Before performing task comorphic partitioning of the processing equipment topology network based on the processing task dataset, the method further includes: Analyze the processing task dataset to obtain the key feature set for task commonality partitioning; The key feature set is divided according to the commonality of the task, and the dimensionality reduction and size reduction processing of the processing task dataset is performed.

5. The distributed control machining method based on CNC precision carving tasks as described in claim 4, characterized in that, The analysis of the processing task dataset obtains a key feature set for task commonality partitioning, including: Based on the principle of standardization, the dataset of the processing tasks is decentralized. Based on the decentralized results, a standard processing data matrix is ​​constructed with the number of processing tasks in the processing task dataset as the number of rows in the matrix and the number of processing index categories in the processing task dataset as the number of columns in the matrix. Establish the correlation matrix of the standard processing data matrix, calculate the eigenvalues ​​and eigenvectors of the correlation matrix, and obtain the feature set of the correlation matrix; Based on the variance contribution scores of the feature set of the correlation matrix, a cumulative variance contribution sequence is obtained by accumulating the variance contribution scores, wherein the cumulative variance contribution sequence includes multiple cumulative variance contribution scores, and the cumulative variance contribution scores have processing feature class labels. Based on the preset contribution rate constraint, the cumulative variance contribution score is filtered according to the cumulative variance contribution sequence, and the processing feature class label of the filtering result is extracted to generate the task commonality partitioning key feature set.

6. The distributed control machining method based on CNC precision carving tasks as described in claim 1, characterized in that, The step of constructing a distributed collaborative processing controller using the processing task dataset and the comorphic partitioning result includes: Randomly select from the common-state partitioning results to obtain the first task common-state partitioning class; Based on the first task common-state classification, and based on the common-state classification result, the first processing data group is invoked; Based on the first set of processed data, a first collaborative processing control strategy is obtained through supervised training. The first collaborative processing control strategy is stored in the distributed collaborative processing controller.

7. The distributed control machining method based on CNC precision carving tasks as described in claim 1, characterized in that, The distributed collaborative control of multiple distributed CNC precision carving machines includes: Analyze the real-time processing status data to determine the common-state classification of real-time tasks; Based on the real-time task common-state classification, the multiple collaborative processing control strategies in the distributed collaborative processing controller are traversed for control matching. Based on the control matching results, the distributed collaborative processing controller is activated to make processing control decisions. The processing equipment topology network distributes the processing control decision results to multiple distributed CNC precision carving processing equipment for distributed collaborative control processing.

8. The distributed control machining method based on CNC precision carving tasks as described in claim 7, characterized in that, The distributed collaborative control process of multiple distributed CNC precision carving machines also includes: Periodically collect data on actual machining dimensions and tool wear from each distributed CNC precision carving equipment, and summarize them to generate a real-time machining error dataset; Calculate the average overall processing error under the current real-time task common-state partitioning class. When the average overall processing error exceeds the preset processing accuracy threshold, re-partition the current processing state data into common-state partitions, re-match the collaborative processing control strategy, and update and issue processing control commands to complete the dynamic correction of processing control.

9. A distributed control machining equipment based on CNC precision carving tasks, characterized in that, The device includes: a memory, a processor, and a distributed control machining program based on a CNC engraving task stored in the memory and executable on the processor, the distributed control machining program based on the CNC engraving task being configured to implement the steps of the distributed control machining method based on a CNC engraving task as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a distributed control machining program based on a CNC precision carving task. When the distributed control machining program based on the CNC precision carving task is executed by a processor, it implements the steps of the distributed control machining method based on a CNC precision carving task as described in any one of claims 1 to 8.

Images