Intelligent manufacturing data processing system based on edge side full-process collaboration

Abstract

The invention discloses an intelligent manufacturing data processing system based on edge side full-process collaboration, and the system comprises an equipment edge node layer which is disposed at each station and logistics node, and each node is integrated with a cache management sub-module, a time sequence calibration sub-module and a signature verification sub-module, and is used for collecting the equipment data of the station, executing cache migration, time sequence calibration, data signature and preorder signature verification; the regional edge gateway layer is used for deploying a station-workpiece association graph engine and a distributed signature account book and circulation path optimization module and is used for storing a station association rule and historical circulation data, pre-judging a workpiece circulation path, issuing a migration instruction, synchronizing signature data of each edge node to construct a distributed traceability account book and optimizing the distributed traceability account book and the circulation path; according to the invention, a core data processing function is deployed on an edge side by using a three-level architecture of the equipment edge node, the regional edge gateway and the cloud management platform, so that the dependence of a traditional'equipment-cloud 'two-level architecture on the cloud is thoroughly eliminated, and the data transmission distance and delay are greatly reduced.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing technology, and in particular relates to an intelligent manufacturing data processing system based on edge-side end-to-end collaboration. Background Technology

[0002] In the field of intelligent manufacturing, data processing systems are the core support for ensuring efficient production operations. Existing technologies generally adopt a full-link architecture of data acquisition, transmission, processing, and storage to achieve real-time acquisition and management of data from various workstations. Most mainstream systems currently consist of three core parts: equipment terminals, data transmission modules, and management platforms. Equipment terminals are responsible for collecting core data such as processing parameters and equipment status, transmitting the data to the management platform via industrial Ethernet, wireless communication, or other methods. Management platforms are typically deployed in the cloud or on local servers, handling data storage, verification, monitoring, and command issuance. To enhance data security, existing technologies often employ symmetric and asymmetric encryption algorithms to encrypt transmitted data. For multi-device collaboration needs, preset calibration rules or basic timing alignment algorithms are used to achieve initial data coordination between different devices. Furthermore, some systems integrate data traceability functions, recording data flow logs to provide basic support for production quality traceability, and are widely used in various intelligent manufacturing scenarios such as automotive manufacturing, electronics processing, and machining.

[0003] While existing data processing systems have achieved basic data management and collaboration functions, they still have many shortcomings that urgently need to be addressed when considering the core requirements of intelligent manufacturing for efficiency, accuracy, security, and flexibility. Existing systems mostly rely on a two-tier "device-cloud" architecture, with core processing functions concentrated in the cloud. This results in long data transmission distances, high latency, and excessive consumption of cloud bandwidth, making it difficult to meet real-time collaboration needs. Data migration often uses conventional transmission protocols without incorporating dedicated optimization technologies and backup mechanisms, leading to improvements in migration efficiency and security. Time-series calibration uses fixed slicing windows or simple algorithms, which cannot adapt to the differences in various types of production data, limiting collaboration accuracy. Data security relies on single encryption or independent signature modes, lacking a chain-like protection system across all stages, making it susceptible to data tampering risks and difficult to accurately locate abnormal links during traceability. Association rules are mostly fixed configurations, unable to dynamically adapt to flexible needs such as workstation adjustments and multi-category production, resulting in insufficient system scalability. Furthermore, most systems lack offline read / write and autonomous collaboration capabilities, making them prone to production stoppages during network outages and unable to meet the diverse production needs of different scenarios. Summary of the Invention

[0004] To overcome the aforementioned deficiencies in existing technologies, this invention provides an intelligent manufacturing data processing system based on edge-side end-to-end collaboration, which solves the problems of insufficient architecture, transmission, calibration, protection, flexibility, and offline adaptation in existing technologies.

[0005] To achieve the above objectives, the present invention provides the following technical solution: The intelligent manufacturing data processing system based on edge-side end-to-end collaboration adopts a three-tier architecture of equipment edge nodes, regional edge gateways, and a cloud management platform. Core functions are deployed at the edge. The system includes: The equipment edge node layer is deployed at each workstation and logistics node. Each node integrates a cache management submodule, a timing calibration submodule, and a signature verification submodule to collect equipment data at its workstation and perform cache migration, timing calibration, data signing, and pre-signature verification. The regional edge gateway layer deploys a workstation-workpiece association graph engine, a distributed signature ledger, and a flow path optimization module. These modules are used to store workstation association rules and historical flow data, predict workpiece flow paths and issue migration instructions, synchronize signature data from each edge node to build a distributed traceability ledger, and dynamically adapt to abnormal flow scenarios. The cloud management platform layer is used for initial configuration and updates of the association graph, backup and synchronization of the distributed ledger, monitoring of the overall system operation status and pushing of abnormal alarms, but does not participate in real-time data collaborative processing. The system is driven by the cross-workstation flow of workpieces, realizing end-to-end data collaboration including flow triggering, dynamic migration, timing calibration, chain signature, and full-chain traceability. The logic is as follows: after the processing of the preceding workstation is completed, the regional edge gateway predicts the target workstation by combining the correlation map and triggers the edge nodes of the preceding workstation to migrate the core data to the target workstation. The migrated data carries timing feature tags, and the target workstation node automatically completes the timing slice alignment with the local data after receiving it. The data generation, migration, and calibration links are solidified through the chain signature of the edge nodes to form a traceable and tamper-proof signature chain.

[0006] Preferably, the workstation-workpiece association graph engine of the regional edge gateway layer constructs an association graph that includes workstation attributes, workpiece type, typical flow path and path probability weight. The initial version is imported from the cloud, and subsequent updates are made by collecting real-time flow data through the regional edge gateway. The path prediction accuracy is no less than 99.5%.

[0007] Preferably, the cache management submodule of the device edge node layer adopts a custom migration protocol based on RDMA optimization to realize cache migration. This protocol supports a transmission rate of 10Gbps, and the migrated data is the core processing data of the workpiece with a volume of less than 10KB after LZ4 compression. The migration process enables a dual-buffer backup mechanism. The preceding node retains the original cache until the migration is confirmed, and the target node creates a temporary buffer area to receive the data. The migration success rate is not less than 99.9%, and the migration timeout threshold is set to 5 milliseconds.

[0008] Preferably, the timing calibration submodule at the device edge node layer performs the following operations: When receiving migration data, it synchronously acquires timing feature labels, which include the data acquisition start timestamp, acquisition period, core device status, and data verification hash value, in JSON format and occupying less than 200B of space; it adjusts the granularity of the slicing window based on the migration data acquisition period, slices the migration data and local device data at the same window granularity, and adds timing feature labels; it calculates the similarity using the core device data slice in the migration data as a benchmark, and if the similarity is not less than 95%, it is determined that the alignment is successful and the dataset is spliced ​​to form a collaborative dataset; otherwise, it triggers secondary calibration.

[0009] Preferably, the granularity of the slice window can be adjusted within the range of 5 to 100 milliseconds. The granularity of the window for processing data is 5 to 20 milliseconds, with a typical value of 20 milliseconds. The granularity of the window for quality inspection data is 5 to 10 milliseconds, with a typical value of 10 milliseconds. The granularity of the window for logistics data is 50 to 100 milliseconds, with a typical value of 80 milliseconds. The secondary calibration method is to modify the granularity of the slice window. The maximum number of secondary calibrations is 2. If it still fails, it will be reported to the regional edge gateway and the abnormal data will be marked.

[0010] Preferably, the signature verification submodule of the device edge node layer integrates the SM2 elliptic curve cryptography algorithm module, which automatically performs the signature operation in the three stages of data generation, migration completion and calibration completion. The signature time is less than 1 millisecond, and the signature timeout threshold is set to 1 millisecond. Each signature includes the node ID, the current data hash value, the signature digest of the previous stage and the precise timestamp, realizing a strong binding between the signature and the data.

[0011] Preferably, the distributed signature ledger of the regional edge gateway layer is built using the LevelDB database, which supports offline reading and writing; the signature chain management module constructs a chain structure according to the workpiece processing sequence and the station flow sequence, and adopts the association mode of previous signature digest plus current signature. If the data or signature of any node is tampered with, subsequent signature verification will fail; the distributed ledger synchronization period is set to 100 milliseconds.

[0012] Preferably, the system's full-process traceability logic is as follows: retrieve the corresponding signature chain through the workpiece's unique identifier, and determine the data integrity by recursively verifying the continuity and legality of each signature in the chain, with a verification delay of less than 50 milliseconds; when data anomalies are detected, locate the workstation where the tampering occurred, the corresponding edge node, and the specific link by comparing the differences in hash values ​​and timestamp order of the signatures of each node in the chain, and at the same time retrieve the backup data in the distributed ledger to complete the recovery.

[0013] Preferably, the device edge nodes adopt industrial-grade edge computing gateways, which integrate encryption chips that support hardware acceleration of the SM2 algorithm, and have the ability to collect data, cache, process and encrypt data at this workstation, adapting to the core requirements of cache migration and data signing.

[0014] Preferably, the regional edge gateway has multi-port aggregation capabilities and data storage and computing capabilities, adapting to the core requirements of flow path prediction and signature chain management; the device edge nodes are equipped with an industrial-grade Linux system, supporting containerized deployment, and adapting to the stable operation of cache management, timing calibration, and signature verification sub-modules; the regional edge gateway is equipped with a Linux system, and the association graph engine is built based on a graph database, adapting to the storage and update requirements of the association graph.

[0015] The technical effects and advantages of this invention's intelligent manufacturing data processing system based on edge-side end-to-end collaboration are as follows: 1. This invention adopts a three-tier architecture of device edge nodes, regional edge gateways, and cloud management platform, deploying core data processing functions on the edge side. This completely eliminates the dependence on the cloud in the traditional two-tier architecture of "device-cloud", significantly reduces data transmission distance and latency, reduces cloud bandwidth usage, avoids production interruptions caused by excessive cloud load or network fluctuations, and significantly improves system stability and real-time performance.

[0016] 2. This invention achieves high-speed, lightweight migration of core data by combining a custom migration protocol based on RDMA optimization with the LZ4 compression algorithm. With the dual-buffer backup mechanism, it ensures the integrity and security of the migrated data, improves the migration success rate, avoids the impact of data loss or migration failure on production flow, and adapts to the data collaboration needs of continuous production across workstations.

[0017] 3. This invention relies on a dynamic time warping algorithm and a flexibly adjustable slice window granularity design to achieve precise time alignment between migrated data and local data, effectively solving the time sequence deviation problem caused by asynchronous operation of multiple devices, ensuring the accuracy of collaborative datasets, providing reliable data support for precision control in subsequent processing, testing and other stages, and improving the overall production quality.

[0018] 4. This invention constructs a chain signature mechanism based on the SM2 elliptic curve cryptography algorithm, combined with a distributed signature ledger, to realize the signature solidification of the entire process of data generation, migration, and calibration, forming an immutable signature chain. This can effectively prevent data from being tampered with, and can also quickly trace the entire life cycle of data through the unique identifier of the workpiece, accurately locate abnormal links and responsible parties, and reduce quality risks and operation and maintenance costs.

[0019] 5. The workstation-workpiece association map of this invention supports dynamic updates and path prediction, which can quickly adapt to different workpiece types, production processes and workstation adjustment needs without reconstructing the system architecture; the equipment edge nodes adopt containerized deployment and scalable interface design, which facilitates the access of new equipment and functional upgrades, and perfectly adapts to the needs of different production modes such as standardization, flexibility and multi-batch production.

[0020] 6. The core functionality of this invention is that edge deployment reduces investment in cloud hardware and maintenance, while the low latency and low bandwidth consumption design reduces network and hardware costs. The distributed architecture, combined with offline read and write capabilities, can ensure the normal operation of core production processes when the network is interrupted. After reconnecting to the network, data is automatically synchronized to reduce production downtime losses, balancing performance improvement and cost control. Attached Figure Description

[0021] Figure 1 This is a flowchart of the intelligent manufacturing data processing system based on edge-side end-to-end collaboration proposed in this invention. Detailed Implementation

[0022] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0023] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include," "contain," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "includes..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0024] refer to Figure 1This invention discloses an intelligent manufacturing data processing system based on edge-side end-to-end collaboration. It adopts a three-tier architecture: equipment edge nodes, regional edge gateways, and a cloud management platform, with core functions deployed at the edge. The system is driven by the cross-workstation flow of workpieces. Equipment edge nodes complete data acquisition, signing, cache migration, and timing calibration. The regional edge gateway predicts the flow path through a workstation-workpiece association graph and synchronizes the signature data of each node to construct a chained signature and a distributed ledger. The cloud is only responsible for association graph configuration, ledger backup, and status monitoring, and does not participate in real-time collaboration. This system achieves end-to-end collaboration from flow triggering, data processing, chained signature, and full-chain traceability, significantly improving the real-time performance, security, and collaborative accuracy of data processing, and adapting to various intelligent manufacturing scenarios.

[0025] General deployment basics: Device edge nodes: Adopting an industrial-grade edge computing gateway, integrating an encryption chip that supports hardware acceleration of the SM2 algorithm, running an industrial-grade Linux system, supporting containerized deployment, and integrating cache management, timing calibration, and signature verification sub-modules.

[0026] Regional edge gateway: It has multi-port aggregation capability and data storage / computing capability, runs on Linux system, and deploys workstation-workpiece association graph engine (based on Neo4j graph database), distributed signature ledger (LevelDB database), and flow path optimization module.

[0027] Cloud management platform: Adopting a B / S architecture, it realizes the initial configuration of the association graph, ledger backup, status monitoring and anomaly alarm, but does not participate in real-time data collaboration.

[0028] Core protocols and algorithms: Cache migration adopts a custom protocol based on RDMA optimization, data compression adopts the LZ4 algorithm, timing calibration adopts the dynamic time warping algorithm, and signature adopts the SM2 elliptic curve encryption algorithm.

[0029] Example 1 This embodiment provides an intelligent manufacturing data processing system based on edge-side end-to-end collaboration for multi-station machining scenarios of automotive engine cylinder blocks (mass standardized production). Specific implementation details include: Purpose of implementation: To address the need for mass standardized production of automotive engine cylinder blocks, this invention solves the problems of high data transfer latency between multiple fixed workstations, insufficient cross-equipment collaboration accuracy, and slow traceability response, and verifies the stability and efficiency of the invention in continuous production scenarios.

[0030] Implementation System: The system is built on a three-tier architecture of "device edge node - regional edge gateway - cloud management platform", with the following specific configuration: (1) Workstation-workpiece association map: includes five workstation attributes: roughing, finishing, drilling, quality inspection, and assembly, engine cylinder block model, unique flow path (roughing → finishing → drilling → quality inspection → assembly), path probability weight 100%, the initial version is imported from the cloud, and no frequent updates are required thereafter.

[0031] (2) Equipment edge nodes: Each workstation is equipped with an industrial-grade edge computing gateway, which integrates an encryption chip that supports hardware acceleration of the SM2 algorithm, is equipped with an industrial-grade Linux system, supports containerized deployment, and integrates cache management, timing calibration and signature verification sub-modules.

[0032] (3) Regional edge gateway: It has the ability to aggregate multiple network ports and data storage / computing capabilities. It is equipped with a Linux system and deploys a workstation-workpiece association graph engine (based on Neo4j graph database), a distributed signature ledger (LevelDB database), and a flow path optimization module.

[0033] (4) Core parameters: The cache migration adopts a custom protocol based on RDMA optimization, the transmission rate is configured to 10Gbps, the size of the migration data after LZ4 compression is ≤8KB, the dual cache backup mechanism is enabled, the migration timeout threshold is 5ms, and the migration success rate is ≥99.9%; the granularity of the time-series calibration slice window is configured according to the work station type, 20ms for processing, 10ms for quality inspection, and 80ms for logistics; the time-series feature label is in JSON format, the space occupied is ≤150B, and the similarity threshold is 95%; the chain signature is automatically executed in the data generation, migration completion, and calibration completion stages, the signature time is <1ms, the signature timeout threshold is 1ms, and the distributed ledger synchronization cycle is 100ms.

[0034] (5) Cloud management platform: It adopts a B / S architecture to realize the initial configuration of the association map, ledger backup, status monitoring and abnormal alarm, but does not participate in real-time data collaboration.

[0035] Implementation steps: (1) System initialization: Complete the initial configuration of the workstation-workpiece association map on the cloud management platform and synchronize the map data to the regional edge gateway; each device edge node completes the deployment of sub-modules and parameter debugging to ensure normal communication with the regional edge gateway.

[0036] (2) First station data processing: After the rough machining station completes the cylinder block machining, the equipment edge node collects core data such as cylinder block size and cutting parameters, automatically performs SM2 signature, uploads the signature data to the regional edge gateway, and caches the core data locally.

[0037] (3) Flow prediction and data migration: After the rough machining station edge node collects the workpiece outbound signal, it synchronizes it to the regional edge gateway; the regional edge gateway predicts the next station to be the finishing station based on the correlation map and issues a migration instruction; the rough machining station edge node migrates the cached data and time sequence feature tags to the finishing station edge node through a custom protocol based on RDMA optimization. After the migration is completed, it receives the confirmation signal from the finishing station and retains the local cache backup until the next stage is completed.

[0038] (4) Timing calibration and signature update: After receiving the migration data, the edge node of the finishing station first verifies the legality of the previous signature, and then starts the timing calibration submodule. The migration data and the data collected by the local processing center are sliced ​​and aligned according to the granularity of the 20ms slice window. The similarity is calculated by the dynamic time warping algorithm to be ≥99%, and the alignment is determined to be successful. The data is then spliced ​​to form a collaborative dataset. Subsequently, a second signature is performed and uploaded to the regional edge gateway to update the distributed signature ledger.

[0039] (5) Multi-station flow iteration: The workpiece is sequentially transferred to the drilling, quality inspection and assembly stations, and steps (3)-(4) are repeated. Each station completes data migration, timing calibration and chain signature.

[0040] (6) Abnormal source tracing and handling: When the quality inspection station finds an abnormal cylinder size, the corresponding signature chain is retrieved through the workpiece's unique identifier to quickly locate the abnormal link; the cloud monitors the operating status of each node in real time and backs up the ledger data synchronously.

[0041] Implementation results: Data loading latency is stabilized at 12-15ms, meeting the production requirement of <20ms; the average daily bandwidth usage of a single production line is reduced by 42%, significantly reducing network resource consumption; the cross-device collaborative verification accuracy reaches 99.2%, ensuring consistent processing precision; the traceability response time is controlled at 35-45ms, significantly improving the efficiency of anomaly location; after 30 days of continuous operation, there are no migration failures or calibration anomalies, and the stability meets the requirements of large-scale continuous production.

[0042] Example 2 This embodiment provides an intelligent manufacturing data processing system based on edge-side end-to-end collaboration for precision machining scenarios of smartphone motherboards (multi-device collaboration, high-precision requirements). Specific implementation details include: Purpose of implementation: To address the challenges of asynchronous operation of multiple devices, large timing deviations, and difficulties in adapting to mixed-model production lines in the precision machining of smartphone motherboards, this invention verifies its adaptability in high-precision, small-batch, multi-model production scenarios.

[0043] Implementation System: Based on a three-tier architecture, the core configuration is as follows: (1) Workstation-workpiece association map: includes four workstation attributes: chip mounting, welding, inspection and packaging, three mainstream motherboard models and corresponding flow paths (chip mounting → welding → inspection → packaging), path probability weight ≥ 99%, and supports dynamic switching of paths according to motherboard model.

[0044] (2) Equipment edge nodes: Each workstation is equipped with an industrial-grade edge computing gateway, which integrates an SM2 encryption acceleration chip, runs an industrial-grade Linux system, and deploys core sub-modules in a containerized manner.

[0045] (3) Regional edge gateway: It has the ability to aggregate multiple network ports and process high-speed data, and deploys a graph database-driven association graph engine, LevelDB distributed ledger and path optimization module.

[0046] (4) Core parameters: The cache migration adopts the 10Gbps RDMA optimized protocol. The migration data includes core data such as chip coordinates and soldering temperature. The size after LZ4 compression is ≤5KB, and the migration timeout threshold is 5ms. The timing calibration slice window granularity is refined to 10ms (chip placement and soldering station), 5ms (inspection station), and 80ms (packaging station). The timing feature label adds precise parameters such as chip placement nozzle pressure and soldering time, with a similarity threshold of 96%. The chain signature includes the motherboard model identifier, and the signature time is <1ms.

[0047] (5) Cloud platform: responsible for the initial configuration of the map and the preset of parameters for multiple models, but does not participate in real-time collaboration.

[0048] Implementation steps: (1) System pre-configuration: Import the associated graphs and calibration parameters corresponding to the three motherboard models into the cloud and synchronize them to the regional edge gateway; each device edge node loads the core sub-module adapted to its workstation and completes the communication debugging with the device.

[0049] (2) Model switching and map adaptation: When producing a certain model of motherboard, the regional edge gateway automatically calls the corresponding flow path map and sends it to the edge nodes of each workstation.

[0050] (3) Data processing at the placement station: After the SMT placement machine completes the placement, the edge node collects and signs data such as placement coordinates and nozzle pressure, and caches them locally.

[0051] (4) Data migration and calibration: The edge gateway of the area triggered by the outbound signal of the patch station predicts that the next station is the welding station and issues a migration instruction; the migration data is transmitted to the welding station node via the RDMA protocol. After the welding station node verifies the signature, it completes the timing alignment with the reflow oven data according to the 10ms slice window, generates a collaborative dataset and signs it a second time.

[0052] (5) Subsequent station transfer: The workpiece passes through the inspection and packaging stations in sequence, and repeats the migration, calibration and signing process. The inspection station improves the calibration accuracy by 5ms slice window.

[0053] (6) Full-process monitoring: Real-time synchronization of production data and ledger backup for each model in the cloud, supporting traceability of model switching records.

[0054] Implementation results: Data loading latency is controlled within 8-12ms, and the cross-device collaborative verification accuracy reaches 99.5%, meeting the high precision requirements of precision machining; traceability response time is <40ms, and the map update time is <1s when switching between multiple models, with no production interruption; the average daily bandwidth usage of a single production line is reduced by 45%, adapting to the flexible needs of small and medium batch multi-model production.

[0055] Example 3 This embodiment provides an intelligent manufacturing data processing system based on edge-side end-to-end collaboration for aerospace component manufacturing scenarios (with high safety and high traceability requirements). Specific implementation details include: Purpose of implementation: To address the stringent requirements of aerospace component manufacturing for data tamper-proofing, high traceability accuracy, and offline operation capabilities, this invention verifies its reliability and compliance in high-security scenarios.

[0056] Implementation System: Built on a three-tier architecture, with enhanced security and traceability design: (1) Device edge node: It adopts a high-protection-level industrial edge computing gateway, integrates SM2 encryption acceleration chip and large-capacity local storage, is equipped with industrial-grade Linux system, and containerizes the core sub-modules to support offline data processing.

[0057] (2) Regional edge gateway: Configure redundant storage and multi-port backup, deploy Neo4j association graph engine, large-capacity LevelDB distributed ledger (supporting 24h offline read and write) and path optimization module.

[0058] (3) Core parameters: Chain signature enhancement design, the signature includes node ID, data hash value, previous signature digest, accurate timestamp and operator ID, SM2 key is automatically updated every 7 days; cache migration adopts "local + gateway" dual backup mechanism, migration data includes key data such as heat treatment temperature curve, LZ4 compressed size ≤9KB, RDMA transmission rate 10Gbps; timing calibration slice window granularity is 15ms (machining station), 20ms (forging, heat treatment station), 10ms (non-destructive testing station), supports offline local alignment.

[0059] (4) Cloud platform: responsible for key management, ledger backup and anomaly alarm, and has the function of synchronization after offline data retransmission.

[0060] Implementation steps: (1) Security initialization: The key distribution and association graph configuration are completed in the cloud, and the regional edge gateway and each device node complete key synchronization and identity authentication.

[0061] (2) Forging station data processing: After the precision forging machine completes the processing, the edge node collects data such as forging parameters and operator ID and signs them, and stores them synchronously to the local and regional edge gateways.

[0062] (3) Offline transfer processing: Simulate network interruption scenario. When the workpiece is transferred to the machining station, the machining node receives the migration data through local cache, completes offline timing calibration and signing, and temporarily stores the data locally.

[0063] (4) Network data synchronization: After the network is restored, each node will automatically synchronize the signature data and calibration data during the offline period to the regional edge gateway to complete the distributed signature chain.

[0064] (5) Non-destructive testing and traceability: When an abnormality is found at the non-destructive testing station, the complete signature chain is retrieved through the workpiece identification to locate the specific equipment and operator.

[0065] (6) Full lifecycle backup: Distributed ledger data is regularly synchronized in the cloud to form multiple data backups.

[0066] Implementation results: The traceability response time is less than 45ms, and the tampering location accuracy reaches specific equipment and operators, meeting the high traceability requirements of aerospace. After 24 hours of offline operation, the data integrity is 100%, and the synchronization success rate after networking is 100%. The data anti-tampering capability has passed industry security certification, and there are no security vulnerabilities or data anomalies after 90 days of continuous operation.

[0067] Example 4 This embodiment provides an intelligent manufacturing data processing system based on edge-side end-to-end collaboration for flexible furniture customization production line scenarios (multi-category, dynamic workstation adjustment). Specific implementation details include: Purpose of implementation: To address the challenges of dynamic workstation adjustments, multi-material compatibility, and unpredictable workflows in furniture customization production, this invention verifies its rapid adaptability in flexible production scenarios, thus avoiding production interruptions caused by the reconstruction of traditional systems.

[0068] Implementation System: Built on a three-tier architecture, emphasizing flexible and adaptable design: (1) Workstation-workpiece association map: Supports dynamic addition / deletion of workstation attributes, and automatically generates flow paths according to furniture material (board, pipe) and order type. The path probability weight is optimized once every 1 hour according to historical orders.

[0069] (2) Device edge node: an industrial-grade edge gateway with scalable interface, equipped with an industrial-grade Linux system, with containerized deployment of core sub-modules, supporting rapid access of new devices.

[0070] (3) Regional edge gateway: It has high-speed data processing capabilities and deploys a dynamically updatable association graph engine, LevelDB ledger and workstation adaptation module.

[0071] (4) Core parameters: The granularity of the timing calibration slice window is adjusted according to the material, 50ms for wood processing and 20ms for metal processing, and automatic parameter synchronization is supported; the cache migration adopts the RDMA optimization protocol, the migration data volume is ≤10KB, and it can adapt to the processing data requirements of different materials.

[0072] (5) Cloud platform: Supports visualization configuration of associated graphs and remote debugging of workstation parameters.

[0073] Implementation steps: (1) Initial deployment: Configure the basic workstation (cutting, grinding, painting, assembly) association map in the cloud and synchronize it to the regional edge gateway; each workstation edge node completes the docking with the corresponding equipment.

[0074] (2) Routine production flow: When board furniture orders enter the production line, the regional edge gateway generates a dedicated flow path, and each workstation completes data migration, calibration and signature according to preset parameters.

[0075] (3) Dynamic workstation adjustment: When a new "carving" workstation is added, the workstation attributes and flow rules are added through the cloud platform and synchronized to the regional edge gateway; the edge node of the carving workstation automatically loads the adaptation parameters and completes the docking with the production line without stopping the machine.

[0076] (4) Multi-material adaptation: When switching to pipe furniture production, the regional edge gateway automatically calls the calibration parameters corresponding to metal processing, and each workstation node synchronously adjusts the granularity of the slice window.

[0077] (5) Effect monitoring: The system operation status after workstation adjustment and material switching is monitored in real time in the cloud, and the adaptation time is recorded.

[0078] Implementation results: Workstation adjustment and adaptation time is less than 30 minutes, without the need to reconstruct the system architecture; data loading latency is stable at 15-20ms, and the accuracy of collaborative verification is 98.5%; it supports production adaptation of multiple materials such as wood and pipes, with a switching time of less than 2 minutes; the operation and maintenance cost of a single production line is reduced by 35%, significantly improving the efficiency of flexible production.

[0079] Example 5 This embodiment provides an intelligent manufacturing data processing system based on edge-side end-to-end collaboration for small-batch, multi-production scenarios of electronic components (high-frequency production changeover, low bandwidth requirements). Specific implementation details include: Purpose of implementation: To address the issues of frequent production changes and limited workshop bandwidth in small-batch, multi-batch production of electronic components, this invention verifies the energy efficiency and stability of low-bandwidth, high-frequency production change scenarios, thereby reducing production energy consumption and network costs.

[0080] Implementation System: Based on a three-tier architecture, bandwidth usage and production switching efficiency are optimized. (1) Workstation-workpiece association map: preset the flow path of 10 electronic components such as resistors and capacitors (stamping → forming → inspection), which can be directly called when changing production without regenerating.

[0081] (2) Device edge nodes: Low-power industrial edge gateways are used, equipped with industrial-grade Linux systems, and core sub-modules are deployed in a containerized manner to optimize data compression and transmission logic.

[0082] (3) Regional edge gateway: Deploy a lightweight association graph engine and LevelDB ledger to support fast path calls.

[0083] (4) Core parameters: Cache migration optimizes LZ4 compression algorithm parameters, increases compression ratio to 12:1, migration data volume ≤3KB, reduces transmission bandwidth usage; chain signature merge migration completion and calibration completion stages, only retains two signature stages: data generation and final calibration completion; RDMA transmission rate is adjusted as needed to adapt to workshop bandwidth limitations of ≤100Mbps.

[0084] (5) Cloud platform: responsible for preset path management and bandwidth usage monitoring.

[0085] Implementation steps: (1) Preset configuration: Import the correlation map and calibration parameters of 10 electronic components into the cloud and synchronize them to the regional edge gateway; each workstation node completes the low power mode configuration.

[0086] (2) High-frequency production changeover preparation: When production changes every 2 hours, the preset path and parameters of the next batch of components are called in advance through the cloud and synchronized to each node.

[0087] (3) Stamping station data processing: After the stamping equipment completes the processing, the edge node collects the stamping parameters and signs them, and then caches them after high compression ratio processing.

[0088] (4) Low-bandwidth migration and calibration: The outbound signal triggers data migration. The migrated data is compressed and transmitted via RDMA protocol. The target workstation node performs the final signature after completing calibration.

[0089] (5) Bandwidth monitoring: The cloud monitors the bandwidth usage in real time and dynamically adjusts the transmission rate to ensure that it does not exceed the workshop bandwidth limit.

[0090] (6) Circular flow: Repeated production change, data processing, migration and calibration process, recording production change time and bandwidth usage.

[0091] Implementation results: Data loading latency is controlled within 10-15ms, meeting the efficiency requirements of small-batch production; the average daily bandwidth usage of a single production line is reduced by 48%, adapting to bandwidth limitations of ≤100Mbps; data collaboration is uninterrupted during production changeover, with a changeover time of <30s; energy consumption is reduced by 20% in low-power mode, significantly reducing production and maintenance costs.

[0092] Comparative Example 1 This comparative example provides a traditional intelligent manufacturing data processing solution, specifically including: Solution design (core differences from the present invention): (1) No edge-side three-level architecture, adopts a "device-cloud" two-level architecture, and the data cache is only stored locally on a single device.

[0093] (2) Data retrieval across workstations requires downloading from the cloud and there is no cached hot migration mechanism.

[0094] (3) Cross-device synchronization relies on cloud-based NTP timing and lacks edge-side time slice alignment.

[0095] (4) Data traceability adopts centralized cloud-based log recording and a non-chain signature mechanism.

[0096] Test conditions: The same automotive engine block machining scenario, equipment, and production cycle as in Example 1.

[0097] Test results: See Table 1 for details.

[0098] Table 1. Comparison of parameters between Comparative Example 1 and Example 1 ; Compared with Examples 1-5 and Comparative Example 1, Comparative Example 1 of the present invention adopts a two-level "device-cloud" architecture, without edge-side collaborative design, data caching is only localized, cross-workstation retrieval relies on cloud download, synchronization relies on cloud NTP time synchronization, and traceability adopts centralized logs. The core has defects such as high latency, large bandwidth consumption, low collaborative accuracy, slow traceability, and poor stability. In contrast, Examples 1-5 are based on a three-level edge collaborative architecture, which achieves comprehensive optimization for different intelligent manufacturing scenarios, and the overall performance far exceeds that of Comparative Example 1.

[0099] In a high-volume standardized production scenario (Example 1), the data loading latency in Comparative Example 1 reached 180-220ms, with an average daily bandwidth consumption of 80GB, a cross-device collaborative verification accuracy of only 82%, and continuous operation without anomalies for only 8-10 days, which could not meet the continuous production requirements of automotive engine blocks. This invention, through cache hot migration and edge timing calibration, reduced the latency to 12-15ms, reduced bandwidth consumption by 42%, improved the accuracy to 99.2%, and achieved stable operation for over 30 days, fully meeting the high efficiency and stability requirements of high-volume production.

[0100] In the precision machining scenario (Example 2), Comparative Example 1, lacking a refined timing calibration mechanism, struggles to adapt to the asynchronous operation characteristics of multiple devices on the mobile phone motherboard, and production is easily interrupted when switching between multiple models. This invention refines the slicing window granularity to 5-10ms, achieving a collaborative accuracy rate of 99.5%, with a multi-model switching pattern update time of <1s and a delay controlled within 8-12ms, precisely meeting the requirements of high-precision, multi-model mixed-line production. Comparative Example 1, however, cannot achieve this kind of precise collaboration.

[0101] In high-security scenarios (Example 3), Comparative Example 1 lacks offline processing capabilities and enhanced security design; network interruption leads to shutdown, and tracing can only locate the device, which cannot meet the requirements of the aerospace industry. This invention supports 24-hour offline read and write, chained signatures include operator IDs, keys are updated regularly, data integrity is 100% after offline operation, tracing can accurately locate the device and personnel, has passed industry security certification, and has been running continuously for 90 days without vulnerabilities. The centralized logs of Comparative Example 1 completely lack this kind of security and offline adaptability.

[0102] In the flexible production scenario (Example 4), the architecture of Comparative Example 1 is fixed, and dynamic workstation adjustments require system reconstruction, resulting in a long adaptation cycle. The association graph of this invention supports dynamic updates, with the adaptation time for adding a new workstation being less than 30 minutes, the time for switching between multiple materials being less than 2 minutes, the collaboration accuracy rate being 98.5%, and the maintenance cost being reduced by 35%. However, Comparative Example 1 cannot cope with the dynamic workstations and multiple material requirements of furniture customization.

[0103] In small-batch, multi-batch scenarios (Example 5), compared to Example 1, which lacks bandwidth optimization, production changeovers require parameter reconfiguration, making it difficult to adapt to workshop bandwidth of ≤100Mbps. The optimized compression algorithm of this invention reduces migration data to ≤3KB, bandwidth usage by 48%, and presets 10 paths to achieve production changeover time of <30s, reducing energy consumption by 20%. It perfectly adapts to high-frequency production changeovers and low bandwidth limitations, while Example 1 suffers from prominent production changeover interruptions and high bandwidth usage.

[0104] In summary, the two-level architecture of Comparative Example 1 has inherent limitations and can only meet basic data processing needs. Examples 1-5, through a three-level edge collaboration architecture, not only solve the common defects of traditional solutions, but also optimize the design for the special needs of different scenarios. It surpasses Comparative Example 1 in all core indicators such as latency, bandwidth, accuracy, traceability, stability, and adaptability, and can cover all sub-segments of intelligent manufacturing.

[0105] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of protection of the claims.

[0106] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent manufacturing data processing system based on edge-side end-to-end collaboration, characterized in that, The system adopts a three-tier architecture consisting of device edge nodes, regional edge gateways, and a cloud management platform. Core functions are deployed at the edge. The system includes: The equipment edge node layer is deployed at each workstation and logistics node. Each node integrates a cache management submodule, a timing calibration submodule, and a signature verification submodule to collect equipment data at its workstation and perform cache migration, timing calibration, data signing, and pre-signature verification. The regional edge gateway layer deploys a workstation-workpiece association graph engine, a distributed signature ledger, and a flow path optimization module. These modules are used to store workstation association rules and historical flow data, predict workpiece flow paths and issue migration instructions, synchronize signature data from each edge node to build a distributed traceability ledger, and dynamically adapt to abnormal flow scenarios. The cloud management platform layer is used for initial configuration and updates of the association graph, backup and synchronization of the distributed ledger, monitoring of the overall system operation status and pushing of abnormal alarms, but does not participate in real-time data collaborative processing. The system is driven by the cross-workstation flow of workpieces, realizing end-to-end data collaboration including flow triggering, dynamic migration, timing calibration, chain signature, and full-chain traceability. The logic is as follows: after the processing of the preceding workstation is completed, the regional edge gateway predicts the target workstation by combining the correlation map and triggers the edge nodes of the preceding workstation to migrate the core data to the target workstation. The migrated data carries timing feature tags, and the target workstation node automatically completes the timing slice alignment with the local data after receiving it. The data generation, migration, and calibration links are solidified through the chain signature of the edge nodes to form a traceable and tamper-proof signature chain.

2. The intelligent manufacturing data processing system based on edge-side end-to-end collaboration as described in claim 1, characterized in that, The workstation-workpiece association graph engine in the regional edge gateway layer constructs an association graph that includes workstation attributes, workpiece type, typical flow paths, and path probability weights. The initial version is imported from the cloud, and subsequent updates are made by collecting real-time flow data through the regional edge gateway. The path prediction accuracy is no less than 99.5%.

3. The intelligent manufacturing data processing system based on edge-side end-to-end collaboration as described in claim 1, characterized in that, The cache management submodule at the device edge node layer uses a custom migration protocol based on RDMA optimization to implement cache migration. This protocol supports a transmission rate of 10Gbps. The migrated data is the core processing data of the workpiece with a volume of less than 10KB after LZ4 compression. The migration process enables a dual-buffer backup mechanism. The preceding node retains the original cache until the migration is confirmed. The target node creates a temporary buffer to receive the data. The migration success rate is no less than 99.9%. The migration timeout threshold is set to 5 milliseconds.

4. The intelligent manufacturing data processing system based on edge-side end-to-end collaboration as described in claim 1, characterized in that, The timing calibration submodule at the device edge node layer performs the following operations: When receiving migration data, it synchronously acquires timing feature tags. The timing feature tags include the data acquisition start timestamp, acquisition period, core device status, and data verification hash value. They are in JSON format and occupy less than 200B of space. Based on the migration data acquisition period, it adjusts the granularity of the slice window, slices the migration data and local device data at the same window granularity, and adds timing feature tags. Based on the core equipment data slices in the migration data, the similarity is calculated using the dynamic time warping algorithm. If the similarity is not less than 95%, it is considered that the alignment is successful and the data is stitched together to form a collaborative dataset; otherwise, a secondary calibration is triggered.

5. The intelligent manufacturing data processing system based on edge-side end-to-end collaboration as described in claim 4, characterized in that, The granularity of the slice window can be adjusted within the range of 5 to 100 milliseconds. The corresponding window granularity for processing data is 5 to 20 milliseconds, with a typical value of 20 milliseconds. The corresponding window granularity for quality inspection data is 5 to 10 milliseconds, with a typical value of 10 milliseconds. The corresponding window granularity for logistics data is 50 to 100 milliseconds, with a typical value of 80 milliseconds. The secondary calibration method is to modify the granularity of the slice window. The maximum number of secondary calibrations is 2. If it still fails, it will be reported to the regional edge gateway and the abnormal data will be marked.

6. The intelligent manufacturing data processing system based on edge-side end-to-end collaboration as described in claim 1, characterized in that, The signature verification submodule at the device edge node layer integrates the SM2 elliptic curve cryptography algorithm module. It automatically performs signature operations in the three stages of data generation, migration completion, and calibration completion. The signature time is less than 1 millisecond, and the signature timeout threshold is set to 1 millisecond. Each signature includes the node ID, the current data hash value, the signature digest of the previous stage, and the precise timestamp, realizing a strong binding between the signature and the data.

7. The intelligent manufacturing data processing system based on edge-side end-to-end collaboration as described in claim 1, characterized in that, The distributed signature ledger of the regional edge gateway layer is built using the LevelDB database and supports offline reading and writing. The signature chain management module constructs a chain structure according to the workpiece processing sequence and the station flow sequence. It adopts the association mode of previous signature digest plus current signature. If the data or signature of any node is tampered with, the subsequent signature verification will fail. The distributed ledger synchronization period is set to 100 milliseconds.

8. The intelligent manufacturing data processing system based on edge-side end-to-end collaboration as described in claim 1, characterized in that, The system's full-process traceability logic is as follows: retrieve the corresponding signature chain through the unique identifier of the workpiece, and determine the data integrity by recursively verifying the continuity and legality of each signature in the chain, with a verification delay of less than 50 milliseconds; when data anomalies are detected, locate the workstation where the tampering occurred, the corresponding edge node, and the specific link by comparing the differences in the hash values ​​of the signatures of each node in the chain and the timestamp order, and at the same time retrieve the backup data in the distributed ledger to complete the recovery.

9. The intelligent manufacturing data processing system based on edge-side end-to-end collaboration as described in claim 1, characterized in that, The device edge nodes adopt industrial-grade edge computing gateways, which integrate encryption chips that support hardware acceleration of the SM2 algorithm. They have the ability to collect data, cache, and perform encryption operations for the device at this workstation, and are adapted to the core requirements of cache migration and data signing.

10. The intelligent manufacturing data processing system based on edge-side end-to-end collaboration as described in claim 1, characterized in that, The regional edge gateway has multi-port aggregation capabilities as well as data storage and computing capabilities, adapting to the core requirements of flow path prediction and signature chain management; the device edge nodes are equipped with an industrial-grade Linux system, supporting containerized deployment and adapting to the stable operation of cache management, timing calibration, and signature verification submodules; the regional edge gateway is equipped with a Linux system, and the association graph engine is built based on a graph database, adapting to the storage and update requirements of the association graph.

Images