Data braiding system based on data grid architecture

Abstract

The invention relates to a data braiding system based on a data grid architecture. The data braiding system comprises a global processing device and a production node processing device used for realizing field autonomy, the global processing device comprises a zero-trust access control module for interacting with external information and a global agent; wherein the received external information is dynamically analyzed by the global agent, and a task of the domain agent allocated to the production node processing device is generated; the production node processing device comprises a dynamic strategy coordination layer which is used for interacting with the global processing device and realizing cross-node access consistency; a plurality of domain agents; and the dynamic strategy coordination layer selects at least one domain agent for processing based on tasks distributed by the global agent. The method has the beneficial effects that the balance of'separate treatment 'and'unification' is realized, the data barrier is broken, the problem of data islands is solved, the data processing time is greatly shortened, and the cross-domain organization coordination cost is saved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial data processing, and in particular to a data weaving system based on a data grid architecture. Background Art

[0002] Over the past two decades, data management has gone through cycles of centralization and decentralization, involving technologies such as databases, data warehouses, cloud data storage, data lakes, data lake houses, data weaves, and data grids.

[0003] The past few years have proven that for most organizations, data is more decentralized than centralized. This results in data being spread across multiple business domains, lacking a unified view and efficient sharing mechanisms. Existing lake warehouse technology relies on ETL batch processing, failing to support real-time data access and decision-making, and limiting its use cases. While existing grid technology achieves distributed management through domain autonomy, it lacks cross-domain functional integration capabilities. Data weaving technology relies on a unified metadata layer to integrate data, but it doesn't fully integrate the governance model of domain autonomy.

[0004] In the field of industrial automation, equipment data (such as PLCs and sensors), production plans (such as ERP and MES), and supply chain data are stored in a decentralized manner. Cross-system queries usually require manual connection, and response delays can reach several hours. Traditional centralized data lakes are difficult to adapt to the autonomy needs of multiple factories / workshops, and cross-domain policy conflicts can lead to more than 30% of production coordination failures. Existing industrial reporting systems rely on ETL batch processing and are unable to support real-time decision-making scenarios such as equipment failure prediction and dynamic scheduling.

[0005] Although existing industrial data platforms (such as SAPMES) can achieve single-factory data integration, cross-factory data sharing requires customized interfaces and high expansion costs; traditional data grid technology lacks the ability to discover semantic associations in industrial scenarios, cannot automatically generate cross-domain views such as equipment-process-order, and has a significant impact on existing systems.

[0006] In summary, in traditional industrial production, data from various factories and workshops is isolated due to heterogeneous systems and inconsistent protocols. Data extraction and integration are time-consuming and labor-intensive, severely hindering production efficiency. Data association and sharing across factories and workshops rely on manual configuration. Given the vast amount of data across factories and workshops, integrating and sharing data consumes significant hardware resources and places heavy computational loads, making it difficult to support real-time decision-making and resulting in overall low efficiency. Therefore, how to quickly and efficiently associate and share data across multiple factories and workshops, providing timely support for enterprise decision-making and planning, has become a pressing technical challenge. Summary of the Invention

[0007] (1) Technical issues to be resolved

[0008] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a data weaving system based on a data grid architecture to integrate data from multiple factories / workshops and achieve fast and efficient association and sharing of data across factories / workshops.

[0009] (2) Technical solution

[0010] In order to achieve the above objectives, the main technical solutions adopted by the present invention include:

[0011] In a first aspect, an embodiment of the present invention provides a data weaving system based on a data grid architecture for use in a process industry comprising at least two production nodes, the data weaving system comprising a global processing device and a production node processing device corresponding to each production node for achieving domain autonomy;

[0012] The global processing device includes: a zero-trust access control module for interacting with external information, and a global agent; wherein the received external information is dynamically parsed by the global agent and generates tasks assigned to the domain agent of the production node processing device;

[0013] The production node processing device includes:

[0014] Dynamic policy coordination layer, used to interact with global processing devices and achieve access consistency across nodes;

[0015] Multiple domain agents;

[0016] The dynamic strategy coordination layer selects at least one domain agent to process the tasks based on the tasks distributed by the global agent.

[0017] An embodiment of the present invention proposes a data weaving system based on a data grid, which combines federal distributed governance with intelligent integration, combines the domain autonomy of the data grid with the intelligent integration of data weaving, and delegates data ownership to the production node processing device (each factory / workshop). At the same time, it achieves cross-domain data access consistency through a dynamic policy coordination layer, supports real-time integration and sharing of multi-source heterogeneous data, achieves a balance between "divide and conquer" and "unification", breaks down data barriers, solves the problem of data islands, and enables the production data, equipment data, quality data, etc. of each production node to be quickly aggregated and integrated in a standardized format, greatly shortening data processing time, significantly reducing cross-domain data query response time, and saving cross-domain organization coordination costs.

[0018] Optionally, the global processing device further includes: a global knowledge graph and a global knowledge graph generation engine;

[0019] The global knowledge graph generation engine is used to dynamically generate a global knowledge graph including supply chain, production and equipment by aggregating semantic layer metadata of the domain knowledge graph through the dynamic strategy coordination layer of each production node processing device;

[0020] Accordingly, the global agent performs dynamic analysis and generates tasks for domain agents assigned to production node processing devices, including:

[0021] The global agent dynamically analyzes external information based on the global knowledge graph and generates tasks for domain agents that are assigned to production node processing devices.

[0022] Optionally, the production node processing device further includes: a domain knowledge graph and a domain knowledge graph generation engine;

[0023] The domain knowledge graph generation engine is used to analyze metadata in each production node processing device in real time and dynamically generate a domain knowledge graph;

[0024] Accordingly, the dynamic strategy coordination layer selects at least one domain agent to process the tasks distributed by the global agent, including:

[0025] The dynamic strategy coordination layer selects at least one domain agent to process the tasks distributed by the global agent according to the domain knowledge graph.

[0026] Optionally, the domain knowledge graph generation engine is used to analyze metadata in each production node processing device in real time and dynamically generate a domain knowledge graph, including:

[0027] The domain knowledge graph generation engine is used to analyze the metadata in each production node processing device in real time, dynamically generate semantic layer metadata using a graph convolutional network model and / or a multimodal deep learning model, and dynamically construct a domain knowledge graph based on the semantic layer metadata.

[0028] Optionally, it is characterized in that

[0029] The production node processing device also includes: an autonomous module for storing and autonomously managing metadata, a stream-batch integration engine;

[0030] The stream-batch integration engine is used to integrate Apache Flink for real-time computing of domain data streams, and integrates DeltaLake to store domain data streams through autonomous modules;

[0031] Accordingly, the domain agent performs computing and / or retrieval tasks based on the domain knowledge graph and the data stream of the stream-batch integration engine.

[0032] Optionally, the multiple domain agents include:

[0033] An extract data agent for extracting metadata related to the task;

[0034] An intelligent device diagnostic agent that integrates LSTM models to analyze device sensor data streams in real time, analyze device energy efficiency, and predict device failures;

[0035] A production scheduling optimization agent that dynamically adjusts production plans based on reinforcement learning and responds to interrupted orders or equipment downtime events;

[0036] AI-based PID parameter self-tuning process parameter tuning intelligent agent;

[0037] A trend prediction agent that predicts future trends of process data based on transfer functions or neural networks.

[0038] Optionally, the dynamic strategy coordination layer selects at least one domain agent for processing based on the domain knowledge graph based on the tasks distributed by the global agent, including:

[0039] The dynamic strategy coordination layer determines the domain agent that performs the task based on the domain knowledge graph, and the domain agent determines the metadata associated with the task based on the domain knowledge graph, and calls the stream-batch integration engine to obtain real-time or historical data for processing.

[0040] Optionally, the global agent is further configured to:

[0041] The task execution results fed back by all production node processing devices are merged and presented to visitors through the zero-trust access control module;

[0042] And / or, the multiple production node processing devices are independent workshops or upstream and downstream devices that have correlated impacts.

[0043] Optionally, the production nodes include: a mechanical manufacturing workshop and a spraying workshop for mechanical spraying of paint; a production node control device is deployed in each workshop, and the production node control device of each workshop obtains production data, monitoring data, operation logs during the production process, and key process parameters in real time to form metadata of the autonomous module;

[0044] The global processing device includes: a general control platform in the enterprise control room or the planning and monitoring area, or a client of each manager's portable device.

[0045] Optionally, the production node control device of the mechanical manufacturing workshop includes:

[0046] An intelligent device diagnostic agent that integrates LSTM models to analyze device sensor data streams in real time, analyze device energy efficiency, and predict device failures.

[0047] When the equipment diagnosis agent performs a task, it determines metadata associated with the task based on the domain knowledge graph, obtains the data stream of the metadata through the stream-batch integration engine for processing, and generates a report on the association between the metadata and the equipment fault;

[0048] The production node control device of the spray painting workshop includes: a data extraction agent for extracting metadata related to the task;

[0049] When the data extraction agent performs a task, it determines the metadata associated with the task based on the domain knowledge graph, obtains the data flow of the metadata through the stream-batch integration engine, and returns it to the global agent through the dynamic strategy coordination layer.

[0050] (3) Beneficial effects

[0051] The beneficial effects of the present invention are: the present invention proposes a data weaving system based on data grid, which combines federal distributed governance and intelligent integration, combines the domain autonomy of data grid with the intelligent integration of data weaving, and delegates data ownership to production node processing devices (each factory / workshop). At the same time, it realizes cross-domain data access consistency through a dynamic policy coordination layer, supports real-time integration and sharing of multi-source heterogeneous data, achieves a balance between "divide and conquer" and "unification", breaks down data barriers, solves the problem of data islands, and enables the production data, equipment data, quality data, etc. of each production node to be quickly aggregated and integrated in a standardized format, greatly shortens data processing time, significantly reduces cross-domain data query response time, and saves cross-domain organization coordination costs.

[0052] Through the deep integration of data grid and data weaving technology, this invention has achieved breakthroughs in industrial data governance, cross-domain collaboration, real-time decision-making, and secure expansion, significantly improving the digital efficiency and intelligence level of process industries, and providing an efficient data infrastructure for scenarios such as smart manufacturing and industrial Internet. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a system architecture diagram of a data weaving system based on a data grid architecture in this embodiment. DETAILED DESCRIPTION

[0054] Before describing the specific implementation manner, the following terms are explained first for ease of understanding.

[0055] The database is designed for transaction processing (OLTP), has strong consistency, supports high-concurrency read and write operations, primarily handles structured data, and stores data in a centralized manner. It is not suitable for unstructured data analysis, large-scale historical data mining, or real-time stream processing. It has difficulty processing petabyte-level data and has high expansion costs.

[0056] Data warehouses are designed for online analytical processing (OLAP), providing centralized, hierarchical, and structured data storage for business intelligence (BI) and complex analytics. They are not suitable for storing unstructured data (logs, videos, etc.), writing real-time data, or conducting exploratory analysis, where ETL processes have high latency and storage costs are high.

[0057] Public cloud data storage utilizes distributed storage technology, virtualization, and networking to provide on-demand scalable storage services. It is not suitable for scenarios with extremely high requirements for data security and privacy, nor for businesses requiring low-latency transaction processing or strong consistency.

[0058] Data lakes allow data to be stored in its original format (structured / semi-structured / unstructured types), and perform data conversion during reading (Schema-on-Read). Therefore, they are not suitable for high-concurrency transaction processing, require real-time analysis with strict data quality, require complex data governance technologies, and are prone to forming "data swamps."

[0059] The Data Lakehouse (integrated data lake and warehouse) combines the advantages of a data lake and a data warehouse, supporting ACID transactions and real-time analytics with a unified storage layer. Its core features are the integration of real-time stream processing and batch processing, and the coexistence of structured governance and flexible storage. It is not suitable for ultra-low-latency transaction processing or scenarios with high data governance and compliance requirements, resulting in extremely high implementation complexity and governance costs.

[0060] Data weaving is an intelligent data management architecture that achieves seamless connectivity and a unified view across heterogeneous data sources through dynamic integration, metadata analysis, and knowledge graph technologies. Its core approach is to leverage AI / ML to automatically discover data connections and provide real-time data governance capabilities. However, its implementation is complex. Poor raw data quality or missing metadata can limit the effectiveness of automated governance. Integration across multi-cloud environments can expose privacy vulnerabilities, and it is not suitable for strong real-time transaction processing.

[0061] A data grid is a decentralized architecture that treats data as a product, autonomously managed by business teams and enabling cross-departmental collaboration through a federated governance framework. This requires breaking down the cultural barriers of data silo thinking. Business teams must assume responsibility for data governance, which can easily lead to conflicting standards and inconsistent implementation, making governance coordination difficult.

[0062] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation methods in conjunction with the accompanying drawings.

[0063] Example 1

[0064] See also Figure 1This embodiment provides a data weaving system based on a data grid architecture for a process industry including at least two production nodes. The data weaving system includes a global processing device and a production node processing device corresponding to each production node for achieving domain autonomy.

[0065] The global processing device includes: a zero-trust access control module for interacting with external information, a global intelligent agent, a global knowledge graph and a global knowledge graph generation engine.

[0066] The zero-trust access control module can use industrial firewall hardware isolation, dynamic token verification, and whitelist access mechanism to achieve zero-trust access and prevent OT domain data leakage.

[0067] Among them, the received external information is dynamically analyzed by the global agent, and tasks of the domain agent assigned to the production node processing device are generated. Specifically, the global agent dynamically analyzes the external information according to the global knowledge graph, and generates tasks of the domain agent assigned to the production node processing device.

[0068] For example, the global agent can build an intent parsing unit based on natural language processing (NLP) to parse user natural language instructions (for example, "Optimize the energy consumption of Factory A"), decompose the task into equipment efficiency analysis and process parameter tuning tasks of Factory A based on the global knowledge graph, and then dynamically assign tasks to relevant domain agents of Factory A based on reinforcement learning.

[0069] The global intelligent agent is also used to merge the task execution results fed back by all production node processing devices and present them to the visitor through the zero-trust access control module.

[0070] The global knowledge graph generation engine is used to aggregate the semantic layer metadata of the domain knowledge graph through the dynamic strategy coordination layer of each production node processing device, and dynamically generate a global knowledge graph including supply chain, production and equipment (for example, raw material inventory fluctuations and order quantity affect production scheduling adjustments).

[0071] See also Figure 1 ,If the global agent does not distribute the task, the task is only executed in the ,global agent, and each production node processing device provides ,the metadata stored by the autonomous module to the global ,processing device through the dynamic policy coordination layer.

[0072] Multiple production node processing devices are independent workshops or upstream and downstream devices that have associated impacts.

[0073] The production node processing device includes: a dynamic strategy coordination layer, multiple domain intelligent agents, a domain knowledge graph, a domain knowledge graph generation engine, an autonomous module for storing and autonomously managing metadata, and a stream-batch integration engine.

[0074] The dynamic policy coordination layer interacts with global processing devices and ensures consistent access across nodes. Specifically, it uses standardized global policies and distribution models based on blockchain smart contracts, custom metadata models (e.g., the Standard Information Model, SIM), or OPC-UA based on the SIM model to achieve consistent access.

[0075] The dynamic policy coordination layer selects at least one domain agent for processing based on the tasks distributed by the global agent. Specifically, the dynamic policy coordination layer selects at least one domain agent for processing based on the domain knowledge graph. More specifically, the dynamic policy coordination layer determines the domain agent to perform the task based on the domain knowledge graph. The domain agent then determines the metadata associated with the task based on the domain knowledge graph and calls the stream-batch integration engine to obtain real-time or historical data for processing.

[0076] Intelligent agents in multiple fields include: data extraction intelligent agents used to extract metadata related to tasks; equipment diagnosis intelligent agents that integrate LSTM models to analyze equipment sensor data streams in real time, analyze equipment energy efficiency, and predict equipment failures; production scheduling optimization intelligent agents that dynamically adjust production plans based on reinforcement learning and respond to insertion orders / equipment downtime events; process parameter tuning intelligent agents based on AI-based PID parameter self-tuning; trend prediction intelligent agents that predict future trends of process data based on transfer functions or neural networks.

[0077] All domain agents can determine the metadata associated with the task based on the domain knowledge graph, that is, extract metadata related to the task. Among them, the data extraction agent is the most basic domain agent. After extracting the metadata related to the task, it obtains the data flow of the metadata from the autonomous module and returns it to the global agent through the dynamic strategy coordination layer for subsequent processing by the global agent.

[0078] It should be noted that this embodiment builds domain agents based on the metadata and domain knowledge graph provided by the autonomous module. In a production node processing device, the number and type of domain agents can be flexibly selected according to actual data processing needs, and are not limited to the above-mentioned contents.

[0079] The domain knowledge graph generation engine is used to analyze metadata from each production node processing device in real time and dynamically generate a domain knowledge graph. Specifically, the domain knowledge graph generation engine is used to analyze metadata from each production node processing device in real time, dynamically generate semantic layer metadata using a graph convolutional network (GCN) model and / or a multimodal deep learning model, and dynamically construct a domain knowledge graph based on the semantic layer metadata.

[0080] It's important to note that metadata refers to basic metadata, such as data about equipment parameters at production nodes. Semantic-layer metadata refers to the relationship between basic metadata and business semantics. For example, the semantics of pump equipment operational quality are related to basic metadata such as flow rate, density, power, and head. The collection of semantic-layer metadata for each production node constitutes domain-level semantics.

[0081] When constructing a domain knowledge graph, the domain knowledge graph generation engine continuously analyzes metadata, automatically discovers data associations, and generates domain-level semantics. For example, a graph convolutional network model automatically discovers association rules between equipment parameters (such as temperature and vibration) and process defects (for example, bearing wear leads to reduced product yield). Combining text, images, and time series data, a multimodal deep learning model achieves cross-modal semantic alignment (for example, the text description of "sensor temperature anomaly" should be close to the corresponding image of temperature curve fluctuations in semantic space, while the text description of "equipment normal operation" should be farther away from the image in semantic space). The multimodal deep learning model can be a contrastive language-image pre-training (CLIP) model or a bimodal interaction model (Visual-Linguistic BERT, ViLBERT). Furthermore, the relationship between domain agents and business semantics is also a form of semantic metadata, typically formed by domain agent registration or solidification, and is used to determine which domain agent will perform a task. Therefore, the domain knowledge graph is used not only to determine task-related metadata, namely equipment parameters, but also to determine the domain agent to be used to perform the task.

[0082] The stream-batch integration engine is used to integrate Apache Flink to calculate domain data streams in real time, and integrate DeltaLake to store domain data streams through autonomous modules.

[0083] Accordingly, the domain agent performs computing and / or retrieval tasks based on the domain knowledge graph and the data stream of the stream-batch integration engine.

[0084] It should be noted that when the production node processing device processes the tasks distributed by the global agent, the dynamic strategy coordination layer first determines the domain agent that performs the task based on the domain knowledge graph. The domain agent then determines the metadata associated with the task, i.e., the device parameters, based on the domain knowledge graph. It then calls the stream-batch integration engine to obtain real-time or historical data for processing. Figure 1 In this embodiment, the domain agent calls the domain knowledge graph through the stream-batch integration engine.

[0085] For each production node processing device, the autonomous module and the dynamic policy coordination layer form a federated data grid architecture. The autonomous module is a domain-autonomous module, responsible for managing the data within each production node, or domain, such as existing data products. This allows each factory / workshop to be an independent data product with data ownership. The dynamic policy coordination layer ensures consistent access across factories / workshops.

[0086] Figure 1 In the figure, production node processing device A and production node processing device B are only architectural diagrams. In practice, more independent autonomous grid nodes can be expanded horizontally, such as adding new factories / workshops, and seamless access can be achieved through the dynamic policy coordination layer to reduce expansion costs.

[0087] A domain knowledge graph generation engine is used within each production node to generate a domain knowledge graph based on the metadata of existing data products. A stream-batch integration engine is built to support domain agents in accessing existing data and domain knowledge graphs through stream data processing and batch data processing to perform computing and retrieval tasks. The dynamic strategy coordination layer provides a unified access interface.

[0088] A global zero-trust access mechanism is used outside the production nodes to control access security. A global knowledge graph generation engine is used to generate a global knowledge graph based on the semantic layer metadata of all production nodes. User intentions are analyzed based on the global knowledge graph, and tasks are split and distributed to corresponding domain agents. Finally, the task execution results are merged.

[0089] This embodiment proposes a data weaving system based on a data grid, which combines federal distributed governance with intelligent integration, combines the domain autonomy of the data grid with the intelligent integration of data weaving, and delegates data ownership to production node processing devices (each factory / workshop). At the same time, it achieves cross-domain data access consistency through a dynamic policy coordination layer, supports real-time integration and sharing of multi-source heterogeneous data, and achieves a balance between "divide and conquer" and "unification" based on the core concept of data weaving, breaks down data barriers, solves the problem of data silos, and enables the production data, equipment data, quality data, etc. of each production node to be quickly aggregated and integrated in a standardized format, greatly shortening data processing time, significantly reducing cross-domain data query response time, and saving cross-domain organization coordination costs.

[0090] This embodiment achieves breakthroughs in industrial data governance, cross-domain collaboration, real-time decision-making, and secure expansion through the deep integration of data grid and data weaving technologies, significantly improving the digital efficiency and intelligence level of process industries, and providing an efficient data infrastructure for scenarios such as smart manufacturing and the Industrial Internet.

[0091] This embodiment constructs a federated data grid architecture, which delegates data ownership to the business domain, namely, each production node processing device. In this embodiment, the global knowledge graph generation engine continuously analyzes the semantic layer metadata of each production node to construct a global knowledge graph for realizing global data association; the global knowledge graph aggregates supply chain-production-equipment data, provides global association analysis (such as the impact of raw material inventory fluctuations on production scheduling), assists management in formulating dynamic strategies, and improves agility in responding to market changes; the zero-trust access control module combines industrial firewall hardware isolation, dynamic token verification and whitelist strategies to prevent OT domain data leakage and meet the high security requirements of industrial scenarios; the global intelligent agent dynamically analyzes user intentions, disassembles tasks and assigns them to domain intelligent agents, and processes tasks of domain intelligent agents. The results are merged and processed again to achieve collaborative operations across factories / workshops without the need for manual customization of interfaces; the dynamic policy coordination layer standardizes global policies and distribution modes to achieve cross-domain access consistency, avoid standard conflicts in traditional grid technologies, and reduce production coordination failures caused by data silos; the domain knowledge graph generation engine continuously analyzes metadata to build a domain knowledge graph, automatically discovers data associations and generates domain-level semantics, reducing the cost of manual data docking; the stream-batch integration engine supports real-time stream data processing and historical batch data storage to improve response speed; domain intelligent agents perform data retrieval and computing tasks, replacing the traditional delayed response mode that relies on human experience.

[0092] Example 2

[0093] This embodiment further explains the present invention based on the first embodiment and in combination with specific application scenarios.

[0094] The system of this embodiment is used in a machinery manufacturing plant that includes a manufacturing workshop and a spray painting workshop. The production nodes include: a machinery manufacturing workshop and a spray painting workshop for spraying paint on machinery; each workshop is equipped with a production node control device, which collects production data, monitoring data, operation logs during the production process, and key process parameters in real time to form metadata for the autonomous module;

[0095] The global processing device includes: a general control platform in the enterprise control room or the planning and monitoring area, or a client of each manager's portable device.

[0096] The production node control devices of the mechanical manufacturing workshop include:

[0097] An intelligent device diagnostic agent that integrates LSTM models to analyze device sensor data streams in real time, analyze device energy efficiency, and predict device failures.

[0098] When the equipment diagnosis agent performs a task, it determines metadata associated with the task based on the domain knowledge graph, obtains the data stream of the metadata through the stream-batch integration engine for processing, and generates a report on the association between the metadata and the equipment fault;

[0099] The production node control device of the spray painting workshop includes: a data extraction agent for extracting metadata related to the task;

[0100] When the data extraction agent performs a task, it determines the metadata associated with the task based on the domain knowledge graph, obtains the data flow of the metadata through the stream-batch integration engine, and returns it to the global agent through the dynamic strategy coordination layer.

[0101] The mechanical manufacturing workshop is responsible for producing machinery, while the painting workshop is responsible for applying coatings to newly produced machinery. Both workshops have deployed independent automated control systems, namely production node control devices, which collect on-site equipment, instrumentation, environmental, and production data. Each workshop uses an in-memory database to provide the latest real-time data from the workshop, a time-series database to record historical trends in key process parameters, and a relational database to store operation logs and abnormal events during the production process. Data between the two workshops is difficult to leverage, making it impossible to use artificial intelligence to analyze problems from a plant-wide perspective.

[0102] The workflow of this embodiment is described using the external message "The impact of the operating status of the manufacturing workshop on the yield rate of the spraying workshop" as an example:

[0103] (1) When receiving the external information "the impact of abnormal motor vibration in the manufacturing workshop on the spraying yield in the spraying workshop", the global intelligent agent splits the external information into "performing vibration data analysis in the manufacturing workshop" and "performing spraying parameter collection in the spraying workshop" tasks according to the global knowledge graph.

[0104] (2) The domain agents of each production node device perform tasks,

[0105] In the mechanical manufacturing workshop, the dynamic policy coordination layer selects an LSTM-based equipment diagnosis agent based on the domain knowledge graph based on the tasks assigned by the global agent. The equipment diagnosis agent then determines the metadata associated with the task based on the domain knowledge graph. This metadata includes real-time time series data from the motor vibration sensor, which is part of the real-time data stream; a historical failure mode dataset (including vibration spectrum characteristics and fault type labels) stored in Delta Lake, which is part of the historical data stream; and equipment metadata, including the motor model, installation location, and maintenance history. The equipment diagnosis agent then invokes a stream-batch integration engine, which uses Apache Flink to calculate vibration spectrum characteristics (including main frequency, amplitude, and harmonic components) in real time. It then processes the task and matches historical failure patterns. The task is processed using the vibration spectrum characteristics and equipment metadata as input, and outputs a failure probability (e.g., 85% probability of bearing wear), a confidence level, and historical similar failure cases. This generates a motor vibration analysis and failure prediction report, which is returned to the global agent through the dynamic policy coordination layer.

[0106] B. In the spraying workshop, the dynamic policy coordination layer selects a data extraction agent based on the domain knowledge graph to perform tasks assigned by the global agent. The data extraction agent uses the domain knowledge graph to determine the metadata associated with the task, including spray pressure, temperature, and yield. The data extraction agent invokes the batch-stream integration engine to obtain the data streams of spray pressure, temperature, and yield, performs correlation analysis, and generates a report on the correlation between spray parameters and yield. This report is then returned to the global agent through the dynamic policy coordination layer.

[0107] (3) The results of the global intelligent agent fusion task are the motor vibration analysis and fault prediction report of the mechanical manufacturing workshop and the spraying parameter and yield correlation analysis report of the spraying workshop, generating a "vibration-spraying parameter-yield" correlation report.

[0108] In this embodiment, for a mechanical manufacturing workshop, the domain knowledge graph includes semantic layer metadata such as the relationship between manufacturing equipment parameters (such as temperature and vibration) and product yield, and the relationship between equipment status (such as normal, faulty, in use, idle, etc.) and equipment parameter fluctuation curves, which are dynamically established and updated based on the graph convolutional network model and multimodal deep learning model. For a spray painting workshop, the domain knowledge graph includes semantic layer metadata such as the relationship between spray equipment parameters (such as spray pressure, spray temperature, spray speed) and product yield.

[0109] The domain agent can call the stream-batch integration engine to achieve unified access to basic metadata in the workshop based on existing real-time database management systems, trend database management systems, log and event database management systems, etc. (for example: access to the equipment temperature change curve or real-time change curve within a certain time interval, access to the operations and abnormal events performed by the equipment within the time interval).

[0110] The equipment diagnosis agent uses the real-time change curve of the motor parameters or the recent historical change curve to match the change curve at the time of fault to predict motor faults based on the LSTM model. The trend prediction agent predicts the future change trend of the motor parameters based on the transfer function or neural network, and feeds it back to the equipment diagnosis agent for fault prediction.

[0111] Autonomous modules store model data, real-time data, trend data, logs, and event data. For example, the autonomous module in the mechanical manufacturing workshop stores status and parameter data for each manufacturing device, sensor output data, and data collection frequency. The autonomous module in the spray painting workshop stores spray pressure, temperature, speed, a list of spray equipment, and their protocol configurations. The data stored in the autonomous modules, the domain knowledge graph, and access methods for domain agents are made accessible to global agents through the workshop's dynamic policy coordination layer using the standard OPC-UA protocol and SIM model.

[0112] The global processing device uses a zero-trust access control mechanism, supports remote access from the external network, and accesses the data and methods of each workshop through the standard OPC-UA protocol and SIM model to monitor the operating status of each workshop. The global knowledge graph is used to establish the relationship between business semantics and workshops. For example, the parameters of equipment such as forging and welding need to be obtained from the manufacturing workshop, and the parameters of equipment such as painting and sandblasting need to be obtained from the spraying workshop. The global agent uses an intelligent dialogue system to analyze user input dialogues based on natural language processing, such as: "The impact of abnormal motor vibration in the manufacturing workshop on the spraying yield of the spraying workshop", and distributes tasks to the domain agents of the corresponding workshops for execution and merges the output execution results, and displays the execution results to the user in an appropriate form. Of course, if there are multiple global agents in the global processing device, the global agent responsible for parsing and distributing tasks can also distribute the split tasks to other global agents.

[0113] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0114] The present invention is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions.

[0115] It should be noted that, in the claims, any reference signs placed between brackets shall not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention may be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In claims enumerating several means, several of these means may be embodied by one and the same hardware. The use of the words first, second, third etc. is for convenience only and does not indicate any order. These words may be understood as part of the component name.

[0116] In addition, it should be noted that, in the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0117] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments after learning the basic creative concept. Therefore, the claims should be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0118] Obviously, those skilled in the art may make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if such modifications and variations fall within the scope of the claims and their equivalents, the present invention shall also include such modifications and variations.

Claims

1. A data weaving system based on a data grid architecture, characterized in that: For a process industry including at least two production nodes, the data weaving system includes a global processing device and a production node processing device corresponding to each production node for achieving domain autonomy; The global processing device includes: a zero-trust access control module for interacting with external information, and a global agent; wherein the received external information is dynamically parsed by the global agent and generates tasks assigned to the domain agent of the production node processing device; The production node processing device includes: Dynamic policy coordination layer, used to interact with global processing devices and achieve access consistency across nodes; Multiple domain agents; The dynamic strategy coordination layer selects at least one domain agent to process the tasks based on the tasks distributed by the global agent.

2. The system according to claim 1, wherein: The global processing device also includes: a global knowledge graph and a global knowledge graph generation engine; The global knowledge graph generation engine is used to dynamically generate a global knowledge graph including supply chain, production and equipment by aggregating semantic layer metadata of the domain knowledge graph through the dynamic strategy coordination layer of each production node processing device; Accordingly, the global agent performs dynamic analysis and generates tasks for domain agents assigned to production node processing devices, including: The global agent dynamically analyzes external information based on the global knowledge graph and generates tasks for domain agents that are assigned to production node processing devices.

3. The system according to claim 1, wherein: The production node processing device also includes: a domain knowledge graph and a domain knowledge graph generation engine; The domain knowledge graph generation engine is used to analyze metadata in each production node processing device in real time and dynamically generate a domain knowledge graph; Accordingly, the dynamic strategy coordination layer selects at least one domain agent to process the tasks distributed by the global agent, including: The dynamic strategy coordination layer selects at least one domain agent to process the tasks distributed by the global agent according to the domain knowledge graph.

4. The system according to claim 3, characterized in that The domain knowledge graph generation engine is used to analyze metadata in each production node processing device in real time and dynamically generate a domain knowledge graph, including: The domain knowledge graph generation engine is used to analyze the metadata in each production node processing device in real time, dynamically generate semantic layer metadata using a graph convolutional network model and / or a multimodal deep learning model, and dynamically construct a domain knowledge graph based on the semantic layer metadata.

5. The system according to claim 3, wherein: The production node processing device also includes: an autonomous module for storing and autonomously managing metadata, a stream-batch integration engine; The stream-batch integration engine is used to integrate Apache Flink to calculate domain data streams in real time, and integrate Delta Lake to store domain data streams through autonomous modules; Accordingly, the domain agent performs computing and / or retrieval tasks based on the domain knowledge graph and the data stream of the stream-batch integration engine.

6. The system according to claim 1, wherein: Multiple domain agents include: An extract data agent for extracting metadata related to the task; An intelligent device diagnostic agent that integrates LSTM models to analyze device sensor data streams in real time, analyze device energy efficiency, and predict device failures; A production scheduling optimization agent that dynamically adjusts production plans based on reinforcement learning and responds to interrupted orders or equipment downtime events; AI-based PID parameter self-tuning process parameter tuning intelligent agent; A trend prediction agent that predicts future trends of process data based on transfer functions or neural networks.

7. The system according to claim 5, characterized in that The dynamic strategy coordination layer selects at least one domain agent for processing based on the domain knowledge graph based on the tasks distributed by the global agent, including: The dynamic strategy coordination layer determines the domain agent that performs the task based on the domain knowledge graph, and the domain agent determines the metadata associated with the task based on the domain knowledge graph, and calls the stream-batch integration engine to obtain real-time or historical data for processing.

8. The system according to claim 1, wherein: The global agent is also used to: The task execution results fed back by all production node processing devices are merged and presented to visitors through the zero-trust access control module; And / or, the multiple production node processing devices are independent workshops or upstream and downstream devices that have correlated impacts.

9. The system according to claim 5, characterized in that The production nodes include: a mechanical manufacturing workshop and a spraying workshop for mechanical spraying paint; each workshop is equipped with a production node control device, which obtains the production data, monitoring data, operation logs and key process parameters of each workshop in real time to form metadata of the autonomous module; The global processing device includes: a general control platform in the enterprise control room or the planning and monitoring area, or a client of each manager's portable device.

10. The system according to claim 9, characterized in that The production node control devices of the mechanical manufacturing workshop include: An intelligent device diagnostic agent that integrates LSTM models to analyze device sensor data streams in real time, analyze device energy efficiency, and predict device failures. When the equipment diagnosis agent performs a task, it determines metadata associated with the task based on the domain knowledge graph, obtains the data stream of the metadata through the stream-batch integration engine for processing, and generates a report on the association between the metadata and the equipment fault; The production node control device of the spray painting workshop includes: a data extraction agent for extracting metadata related to the task; When the data extraction agent performs a task, it determines the metadata associated with the task based on the domain knowledge graph, obtains the data flow of the metadata through the stream-batch integration engine, and returns it to the global agent through the dynamic strategy coordination layer.

Images