Numerical control equipment working data acquisition and processing platform

Abstract

The invention discloses a working data acquisition and processing platform for numerical control equipment. The platform is composed of a data acquisition module, a data storage module, a data calling module, an equipment modeling module and a data visualization module. The platform realizes efficient and credible acquisition of multi-source heterogeneous data through a multi-protocol plug-in, edge computing and an event-driven mechanism; data association, anomaly recognition and real-time early warning are completed based on semantic modeling and a graph neural network; ledger management, alarm linkage and a unified data service interface are integrated, and multi-system butt joint and business expansion are supported; and the visualization module supports dynamic display of the equipment state, the processing process and the early warning information. According to the method, the problems of limited data acquisition dimension, weak intelligent analysis capability, poor system expansibility and the like of an existing acquisition platform are solved, the method is suitable for complex state monitoring and predictive maintenance of production equipment in industrial scenes of high-end equipment manufacturing, intelligent workshop management and the like, and the method has good universality and popularization value.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology, specifically to a platform for acquiring and processing working data of CNC equipment. Background Technology

[0002] With the rapid development of intelligent manufacturing and the Industrial Internet, the automation level of CNC equipment in the production process is constantly increasing, thus placing higher demands on the real-time perception, intelligent analysis, and refined control of equipment operating status. As the foundation of CNC equipment information management, the performance of the data acquisition platform directly affects subsequent key aspects such as equipment status monitoring, fault early warning, production quality analysis, and optimization. However, existing CNC equipment data acquisition platforms still have many shortcomings in practical applications, making it difficult to meet the intelligent needs of complex industrial scenarios.

[0003] First, existing data acquisition systems are relatively limited in the scope and type of data collected, mainly focusing on basic physical parameters such as temperature, current, and rotational speed. They neglect other key data types, including program running status, alarm logs, communication information, processing logic, and data quality indicators. This results in insufficient understanding of the equipment's operating status, hindering multi-dimensional analysis and precise control. Furthermore, existing platforms only provide a superficial understanding of the collected data through recording and visual display, failing to construct semantic-level mappings between equipment structure, operating logic, and status behavior. This lack of cognitive ability regarding equipment behavior prevents the interpretation of data within a business context, severely restricting the system's intelligent reasoning and... In terms of decision support capabilities, traditional platforms typically employ threshold-based rule-based judgments or single-point parameter monitoring, making it difficult to identify coupling relationships and hidden risks between variables. They are unable to predict trend anomalies or identify multi-factor collaborative faults, resulting in high false alarm rates and missed alarm risks. Traditional platforms are closed in terms of data services, with weak interface capabilities, lacking a unified data calling mechanism and flexible query capabilities, which is not conducive to cross-system data reuse and rapid expansion of business functions. The system architecture is mostly custom-developed, with high module coupling, making expansion and upgrades difficult and unable to adapt to the development needs of advanced technologies such as edge computing, artificial intelligence algorithms, or large industrial models.

[0004] In summary, existing technologies have significant shortcomings in terms of data acquisition and coverage, equipment semantic understanding, intelligent analysis capabilities, system openness, and functional scalability. A new CNC equipment data acquisition and management solution with a clear structure, complete functions, and higher level of intelligence is needed to meet the actual needs of the next generation of intelligent manufacturing systems. Summary of the Invention

[0005] The objective of this invention is to provide a CNC equipment working data acquisition and processing platform to solve the problems mentioned in the background art.

[0006] The platform is built with a modular architecture and includes five modules: data acquisition module, data storage module, data retrieval module, device modeling module, and data visualization module.

[0007] The data acquisition module is designed to adapt to the industrial communication protocols of different brands of CNC equipment. Within the system, various industrial communication protocols are encapsulated as independent protocol plug-in modules. Each protocol module has independent driver logic, connection mechanisms, and data parsing rules, and is loaded on demand within the platform according to actual needs. These plug-ins are registered to the protocol manager through a unified interface standard, ensuring decoupling and compatibility with the main acquisition system. When a new protocol or equipment model is added, only a new plug-in needs to be developed and introduced, without modifying the core acquisition logic of the existing system. The protocol plug-ins are hot-swappable, allowing protocol drivers and rule configurations to be added, updated, and uninstalled in real time during system operation without stopping the service or restarting the system, ensuring uninterrupted acquisition service.

[0008] Furthermore, to ensure the authenticity, reliability, and verifiability of the collected data and to prevent human tampering or sensor deception: an industrial-grade edge gateway device with an integrated TPM chip is deployed at the forefront of data collection; after each piece of collected data is generated within the edge gateway, a data signature mechanism is used to add a timestamp, device ID, and digital signature to the collected data; a lightweight blockchain technology (Hyperledger Fabric) is deployed within the platform to hash the key raw data to form a data digest, which is then recorded on the blockchain, while the raw collected data is stored in a local or cloud database.

[0009] Furthermore, to ensure the real-time nature of the acquired data, an event-triggered hybrid sampling mechanism is adopted: In addition to conventional timed polling, an event detection subsystem is added. Within this subsystem, pre-set trigger conditions for "critical events" such as changes in vibration, current, or digital input signals are used. Once an event is detected, the high-frequency sampling mode is immediately switched, while simultaneously window-re-sampling of past low-frequency data is performed to achieve complete alignment of data before and after the event. The acquired data is analyzed and judged locally at the edge, prioritized, and the key data is pushed to the central system in real-time using the MQTT protocol. The original data is completely preserved for subsequent traceability and analysis.

[0010] Furthermore, to ensure the integrity of the collected data, an intelligent data collection guardian agent is deployed at the edge collection layer. This component is embedded in each edge node and monitors the collection frequency, data fluctuation trends, and collection stability in real time based on an anomaly detection model. Once potential anomalies such as sensor failure, data interruption, or buffer overflow are detected, the system will immediately issue an alert or automatically restart the collection service, thereby achieving proactive protection. To further address unavoidable extreme packet loss or interruption situations, a pre-trained time-series prediction model LSTM is used to estimate and complete missing data segments. The completed results are clearly labeled and distinguished from the original collected data to ensure that the source is not confused in subsequent analysis, visualization, or tracing scenarios.

[0011] Furthermore, to enable rapid identification and accurate early warning of abnormal states of CNC equipment during data acquisition, the platform constructs a semantic-driven alarm engine + graph neural network anomaly detection mechanism, and continuously uses the acquired data to train and improve the graph neural network (GNN) analysis engine to identify complex interaction patterns and weak signal anomaly combinations hidden behind multi-source data, thereby enabling early prediction of faults.

[0012] Furthermore, in terms of alarm response and communication mechanisms, a multi-channel alarm push system has been established, including SMS, voice calls, App push, email, etc., and the alarms are automatically categorized and pushed according to the urgency and business impact level of the self-built events.

[0013] Furthermore, to ensure system security, a device identity whitelist and AI dynamic risk assessment are set up, and system permissions are restricted based on RABC (role-based access control).

[0014] The data storage module has an underlying edge cache layer deployed near the acquisition terminal to receive and cache the raw data uploaded by the acquisition device.

[0015] Furthermore, data is uniformly accessed and piped through a real-time stream processing layer. This layer employs a distributed message queue middleware to control the order of data streams, mark their sources, and ensure reliable delivery, while also supporting data replay and error retransmission. The real-time processing framework performs pre-cleaning, verification, and tagging on the incoming data, laying a foundation for data consistency for subsequent storage and retrieval.

[0016] Furthermore, the third layer is a hot-cold separation storage layer, which is the core data storage area in the entire system. The hot data section is responsible for storing recently accessed high-frequency data, using a high-performance time-series database or columnar database for storage; cold data is stored in a distributed file system through archiving and compression, used for historical tracing, long-term analysis, and AI model training. In the hot-cold separation strategy, the system dynamically identifies data access hotspots and automatically migrates cold data back to the hot zone to achieve a "cold-to-hot" mechanism, thereby improving response speed and system efficiency.

[0017] Furthermore, the topmost layer is the semantic data layer, dedicated to storing alarm event chains, equipment operation graphs, and knowledge relationships. This layer is built upon a graph database and a semantic indexing engine, providing fundamental support for predictive maintenance, fault tracing, and intelligent search.

[0018] The data retrieval module adopts a unified gateway-driven multi-interface retrieval architecture. Through modular encapsulation and intelligent push mechanism, it enables efficient, flexible, and secure access to different types of industrial data, providing data support for production site visualization systems, MES, and upper-level enterprise information systems.

[0019] The core of the entire data access system is a unified data service gateway. This gateway encapsulates various types of data interfaces on the system bus, including real-time data interfaces, historical data interfaces, alarm event interfaces, and data export interfaces for AI training and BI analysis. All calls are authenticated, routed, and responded to through this gateway, ensuring controllable access permissions, unified and auditable interface calls. The interface design adopts standardized REST and WebSocket protocols, providing clearly structured request parameters and response formats.

[0020] Furthermore, to meet the big data access needs of AI training and BI analysis, the system supports batch export of structured data, allowing users to set sampling windows, data labels, and export formats as needed. This approach provides accurate input for AI models and supports the dynamic generation of various industrial reports and operational dashboards.

[0021] The equipment modeling module, based on the acquisition of equipment operation data, enables the platform to achieve comprehensive digital management of CNC equipment by constructing a unified equipment digital model and ledger management system.

[0022] First, the platform establishes a structured data model for each device using standardized modeling methods. This model covers static attributes such as device number, type, control system, production line, supported processes, communication protocols, and interface parameters. It also defines the device's data acquisition capabilities, including the types of operating parameters, physical units, sampling frequency, and alarm thresholds. All device models have universal templates and personalized expansion capabilities, supporting flexible expansion for subsequent integration.

[0023] Furthermore, based on the structured modeling of equipment, the system synchronously establishes an equipment ledger, assigns a unique digital identity ID to each piece of equipment, and records its entire lifecycle information, including procurement information, commissioning records, production date, maintenance history, parts replacement records, and maintenance plans. At the same time, it is deeply integrated with modules such as real-time operating data, alarm records, and energy consumption analysis to form a complete chain from static description to dynamic monitoring.

[0024] Furthermore, the equipment modeling platform can dynamically assess equipment status, calculate the Equipment Health Index (EHI), and support intelligent maintenance reminders based on runtime or status. Simultaneously, the system supports mobile scanning to view equipment files, quickly locate operational data, alarm history, and maintenance recommendations, improving fault response efficiency and asset management accuracy.

[0025] The data visualization module features a professional large-screen dashboard that collects data on the operating status of the application monitoring system, enabling remote viewing and control. The large screen supports real-time refresh, automatic slideshow, and interactive clicks, and is compatible with touchscreens or remote large-screen control. The dashboard is laid out according to function and data hierarchy, highlighting key equipment status information and anomaly warnings.

[0026] This invention provides a platform for acquiring and processing working data of CNC equipment. Compared with the prior art, it has the following advantages:

[0027] This CNC equipment data acquisition and processing platform has comprehensive advantages in terms of systematization and intelligence, which are reflected in several aspects: First, at the data acquisition level, the platform expands the acquisition types to multiple dimensions such as program execution status, energy consumption information, alarm records, communication status and data quality, based on the traditional mode of only collecting basic operating parameters such as temperature and current. This constructs a comprehensive digital profile of equipment operation and realizes refined control and dynamic monitoring of equipment status.

[0028] At the understanding and decision-making level, the platform deploys device semantic models to perform semantic association modeling on various types of collected data, enabling the system to "understand" the business meaning of each data point and form a semantic-driven rule engine and decision-making basis. At the same time, it constructs complex topological associations of device status through graph neural networks, discovers implicit connections between parameters, and achieves intelligent improvement from anomaly detection to trend prediction and causal analysis.

[0029] In terms of technical architecture, the platform adopts a highly modular design, encompassing sub-modules such as edge acquisition, data platform, semantic engine, GNN graph analysis, visualization dashboard, alarm linkage, and unified interface services. This ensures both decoupling and stability, while also providing flexibility for functional iteration and multi-device expansion. The alarm system features a dual recognition mechanism driven by semantic rules and based on graph intelligence, supporting multi-channel push notifications and tiered early warning systems to achieve high-precision, low-error linkage responses.

[0030] In addition, the system has an efficient data scheduling mechanism that supports on-demand access, tag-based querying, multidimensional indexing, and API calls, meeting the needs of enterprises at all levels for efficient data utilization.

[0031] This CNC equipment data acquisition and processing platform can effectively reduce equipment failure rates, improve production efficiency, and optimize the operation and maintenance cost structure. Through professional visual dashboards, it ensures the stability of system operation and the transparency of management decisions. At the same time, the platform has broad protocol compatibility and intelligent model self-learning capabilities, and has high forward-looking and sustainable evolution potential, making it an important infrastructure for realizing intelligent manufacturing. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the data acquisition and processing platform structure. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] The following example, using the production of precision structural components for the aerospace field, illustrates the implementation process of this invention.

[0035] In this embodiment, for the production of this batch of precision structural parts, the platform access equipment includes CNC horizontal honing machine, CNC lathe, CNC gantry milling machine, magnetic particle flaw detector, and testing equipment. The platform collects data through the communication gateway of the CNC equipment, deployed external sensors, and API interfaces. The data types and contents collected include:

[0036] 1. Basic equipment status data, including whether the equipment is currently powered on, running, or in standby, paused, or faulty status, and its operating mode, such as automatic, manual, or debugging, to understand the current control logic and behavioral environment of the equipment.

[0037] 2. Equipment control and program-related data, including the currently executing machining program number, program segment number, the tool number and position being used, and G-code and M-code instructions received by the equipment. This data reflects the internal operating logic and execution status of the CNC system and is used for process analysis, machining path optimization, and program traceability.

[0038] 3. Key operating parameters of the equipment, including real-time spindle speed, load ratio, feed rate, rapid traverse speed, tool wear, spindle temperature, clamping pressure, etc. These indicators are used for process monitoring, providing basic data for equipment health diagnosis and process optimization.

[0039] 4. Energy consumption and resource usage data, such as the electrical energy consumption, compressed air consumption, coolant usage, and lubricating oil condition generated during equipment operation. This data is used to assess unit energy consumption and achieve energy-saving analysis.

[0040] 5. Equipment health and fault data: Collects various alarm codes and descriptions, alarm time, historical alarm frequency, vibration values, voltage fluctuations, current anomalies, overload or overtemperature events, etc. This data is used for real-time alarms and can also be analyzed to determine the root cause of faults and predict equipment degradation trends.

[0041] 6. Environmental and interference data, including workshop temperature and humidity, airborne dust or oil mist concentration, vibration and noise levels around the equipment, and auxiliary data such as access control records for personnel entering the equipment area. These are used to assess the impact of the external environment on equipment stability.

[0042] 7. Data related to production operations, such as work order numbers, product models, quantity of processed parts, processing time, shift information, yield rate, and defect codes. This type of data is obtained by connecting to MES or ERP systems and, after integration with on-site data, enables traceability and analysis of the entire production process.

[0043] In this embodiment, the CNC equipment working data acquisition and processing platform is constructed in a modular manner, comprising five major modules, the structure of which is shown in the attached figure. Figure 1 As shown, these are the data acquisition module, data storage module, data retrieval module, equipment modeling module, and data visualization module, respectively.

[0044] The data acquisition module is designed to adapt to various industrial communication protocols of CNC equipment from different brands, such as FANUCFOCAS, Modbus, OPCUA, MTConnect, EtherCAT, and Siemens S7. Within the system, these protocols are encapsulated as independent protocol plug-in modules. Each module has its own independent driver logic, connection mechanism, and data parsing rules, which are loaded on demand within the platform. These plug-ins are registered to the protocol manager through a unified interface standard, ensuring decoupling and compatibility with the main acquisition system. When a new protocol or device model is added, only a new plug-in needs to be developed and introduced; the core acquisition logic of the existing system does not need to be modified. The protocol plug-ins are hot-swappable, allowing protocol drivers and rule configurations to be added, updated, and uninstalled in real time during system operation without stopping the service or restarting the system, ensuring uninterrupted acquisition service.

[0045] To ensure the authenticity, reliability, and verifiability of the collected data and prevent human tampering or sensor deception, an industrial-grade edge gateway device integrating a TPM chip is deployed at the forefront of data collection. This ensures at the hardware level that the data collection program running locally on the device and the data it generates cannot be maliciously tampered with or forged by third parties or the operating system itself. After each piece of collected data is generated within the edge gateway, a data signature mechanism is used to add a timestamp, device ID, and digital signature to each piece of collected data, allowing for data traceability and verification at any time during data transmission. To further enhance the immutability of important data, the platform deploys lightweight blockchain technology (Hyperledger Fabric). Key raw data is hashed to form a data digest, which is recorded on the blockchain. The raw collected data is stored in a local or cloud database and is not directly uploaded to the blockchain to reduce performance burden and storage costs. Each evidence storage operation in the blockchain has a chain structure and consensus mechanism to ensure that it cannot be tampered with or deleted later.

[0046] To ensure real-time data acquisition, an event-triggered hybrid sampling mechanism is employed. In addition to conventional timed polling, an event detection subsystem is added. Event trigger conditions are pre-set based on changes in vibration, current, or digital input signals, defining "critical events" such as tool change initiation, interruption alarm, and sudden temperature rise. Once an event is detected, the system immediately switches to a high-frequency sampling mode while simultaneously re-sampling past low-frequency data to ensure complete alignment of data before and after the event. The edge's local caching and analysis capabilities are utilized to locally assess the acquired data, identifying issues such as abnormal vibration and timeout shutdowns. Critical data is prioritized and immediately pushed to the central system, while the original data is fully preserved for subsequent traceability and analysis. Data is transmitted from the edge to the central system using the MQTT protocol.

[0047] To ensure data integrity, an intelligent data acquisition guardian agent is deployed at the edge acquisition layer. This component is embedded in each edge node and monitors the acquisition frequency, data fluctuation trends, and acquisition stability in real time based on an anomaly detection model. Upon detecting potential anomalies such as sensor failure, data interruption, or buffer overflow, the system immediately issues an alert or automatically restarts the acquisition service, thus achieving proactive protection. A multi-path redundant acquisition mechanism is employed, simultaneously acquiring the same type of data from the main PLC and MQTT wireless sensors. This ensures that if one path fails, the system still has a backup path to obtain data, avoiding data loss caused by single-point failures. To further address unavoidable extreme packet loss or interruption situations, a pre-trained time-series prediction model (LSTM) is used to estimate and complete missing data segments. The completed data is explicitly labeled and distinguished from the original acquired data, ensuring that the source is not confused in subsequent analysis, visualization, or tracing scenarios.

[0048] To achieve rapid identification and accurate early warning of abnormal states in CNC equipment during data acquisition, the platform constructs a semantic-driven alarm engine combined with a graph neural network anomaly detection mechanism. The construction process includes:

[0049] 1. Equipment structure modeling (static semantics)

[0050] Define the physical structure hierarchy and component relationships of the equipment. For a CNC machine, it is necessary to model its key components, such as:

[0051] Main structural hierarchy: CNC system → control module → spindle, motor, servo system, cooling unit, sensor;

[0052] The properties of each component include: such as the spindle's rated speed, allowable temperature range, and current characteristics.

[0053] Establish a relationship graph between "components, attributes, units, and constraints".

[0054] This step enables the system to have a "structural understanding" of the device, which facilitates subsequent component-based state monitoring and event modeling.

[0055] 2. State and Behavior Modeling (Dynamic Semantics)

[0056] Define the possible state modes that the device may exhibit during operation and their transition logic, for example:

[0057] Status nodes: Power on, Self-test, Idle, Processing, Alarm, Maintenance in progress;

[0058] Behavioral events: Start machining → Load program → Start spindle → Monitor machining → End machining;

[0059] Condition chain: The processing status must meet the following conditions: "program is valid + cooling is on + workpiece is clamped".

[0060] These states, events, and conditions are described using an event flow diagram (EPC) and stored in the semantic engine, enabling the system to understand "what is normal processing" or "what is an inconsistent state".

[0061] 3. Multi-source data label association (data semantics)

[0062] Bind semantic tags to data streams from multiple sources such as PLCs, sensors, and MES systems, for example:

[0063] Spindle speed > 0 and current > threshold = "Spindle is machining";

[0064] Hydraulic pressure > normal range + duration > 5 seconds = "hydraulic system malfunction";

[0065] An increase in the rate of temperature change, accompanied by increased vibration, equals the "risk of thermal expansion and vibration coupling".

[0066] This step is crucial for linking physical signals with business semantics, enabling the system to have "semantic awareness" capabilities.

[0067] 4. Business rule abstraction (semantic rule engine)

[0068] Various "business scenarios" and "abnormal logic" are described through graphical modeling or a rule-based DSL (Domain Specific Language), forming a semantic rule chain of "if...then...", such as:

[0069] If: during machining + spindle current fluctuations + continuous temperature increase + insufficient coolant flow.

[0070] Therefore, the risk of spindle cooling failure is identified, and a warning is issued.

[0071] These semantic rules can be used in combination and executed in real time through engine parsing.

[0072] 5. Model Visualization and Continuous Optimization

[0073] The completed device semantic model can be displayed graphically in the system, allowing maintenance personnel to: visually view the current status of the device and the semantic tag matching results; review or supplement new rules and logic; introduce expert knowledge or historical cases to continuously train and optimize the model; and support iterative enhancement of the model to continuously improve the accuracy of understanding and the breadth of application.

[0074] The platform further uses the collected data to train and improve the graph neural network (GNN) analysis engine, identifying complex interaction patterns and weak signal anomalies hidden behind multi-source data. For example, when three seemingly minor signal combinations appear, such as "increased spindle temperature rise rate + gradually enhanced vibration + fluctuating machining load", the GNN can determine that they are precursors to "potential spindle bearing failure", thus enabling early prediction of failure.

[0075] Regarding alarm response and communication mechanisms, a multi-channel alarm push system has been established, including SMS, voice calls, App push notifications, and emails. Alarms are automatically categorized and pushed based on the urgency and business impact of self-built events. For example, minor fluctuations are recorded in system logs, intermediate anomalies are alerted via the App, and major hidden dangers or downtime risks are pushed to the responsible personnel via emergency phone and voice linkage, achieving tiered response and rapid handling.

[0076] To ensure system security, a device identity whitelist and AI dynamic risk assessment are set up, and system permissions are restricted based on RABC (role-based access control).

[0077] The data storage module firstly includes an edge caching layer deployed near the acquisition terminal to receive and cache the raw data uploaded by the acquisition device. This layer prioritizes retaining important event data such as temperature anomalies and key indicators such as mechanical shocks.

[0078] Next, the data is uniformly accessed and piped through the real-time stream processing layer. This layer employs a distributed message queue middleware to control the order of the data stream, mark its source, and ensure reliable delivery, while also supporting data replay and error retransmission. The real-time processing framework performs pre-cleaning, validation, and tagging on the incoming data, laying a foundation for data consistency for subsequent storage and retrieval.

[0079] The third layer is the hot / cold tiered storage layer, which is the core data storage area in the entire system. The hot data section stores frequently accessed data, such as equipment status and operating indicators for the past one to three months, using a high-performance time-series or columnar database to support rapid aggregation and trend queries. Cold data is archived and compressed into object storage or a distributed file system, primarily used for historical tracing, long-term analysis, and AI model training. In this hot / cold tiered strategy, the system dynamically identifies data access hotspots and automatically migrates cold data back to the hot zone, implementing a "cold-to-hot" mechanism to improve response speed and system efficiency.

[0080] The top layer is the semantic data asset layer, dedicated to storing alarm event chains, equipment operation graphs, and knowledge relationships. This layer is built on a graph database and a semantic indexing engine, providing fundamental support for predictive maintenance, fault tracing, and intelligent search. For example, when a temperature anomaly alarm occurs, the system not only records the alarm time and equipment status, but also stores the associated operating parameters, historical trends, and possible root causes of the fault.

[0081] The above data storage design can support basic data collection and recording, and also provides a flexible data support system for upper-level monitoring visualization, alarm linkage, AI prediction and report analysis.

[0082] The data retrieval module in this CNC equipment data acquisition system adopts a unified gateway-driven multi-interface retrieval architecture. Through modular encapsulation and intelligent push mechanism, it realizes efficient, flexible and secure access to different types of industrial data, providing data support for production site visualization systems, MES, and upper-level enterprise information systems.

[0083] The core of the entire data access system lies in the unified data service gateway. As the system bus, it encapsulates various types of data interfaces, including real-time data interfaces, historical data interfaces, alarm event interfaces, and data export interfaces for AI training and BI analysis. All calls are authenticated, routed, and responded to through this gateway, ensuring controllable access permissions, unified and auditable interface calls. To reduce learning and development costs, the interface design adopts standardized REST and WebSocket protocols and provides clearly structured request parameters and response formats, enabling callers to obtain the required data with minimal integration overhead.

[0084] For real-time data retrieval, the system integrates edge caching, device push, and WebSocket subscription mechanisms. Data retrieval does not rely on traditional high-frequency polling; instead, it uses an intelligent triggering strategy based on device status, pushing data only when critical status changes occur, thus significantly reducing system load. The real-time data stream can also be directly linked to alarm systems or visualization interfaces via subscription, achieving millisecond-level response monitoring.

[0085] For historical and trend data retrieval, the system is built on a high-performance time-series database and incorporates an intelligent aggregation and completion mechanism. Users can specify the time interval, sampling granularity, and aggregation method, and the system automatically handles time alignment and missing data imputation to ensure complete and consistent query results. This mechanism not only improves query speed but also simplifies the client's data post-processing workflow.

[0086] Regarding alarm and event chain invocation, the system employs a graph database to construct an alarm cause-effect graph and supports natural language-level event queries through a semantic interface. For example, users can query using keywords such as "spindle overheating causing shutdown," and the system will automatically return the context state chain and potential root causes of the corresponding alarm. This design significantly improves the understanding and operational efficiency for non-technical users.

[0087] Furthermore, to meet the big data access needs of AI training and BI analysis, the system supports batch export of structured data, allowing users to customize sampling windows, data labels, and export formats. This not only provides accurate input for AI models but also supports the dynamic generation of various industrial reports and operational dashboards.

[0088] To ensure system performance and stability, the system integrates multi-tenant isolation strategies and call log auditing, enabling the construction of an efficient and controllable data call module.

[0089] The equipment modeling module, based on the completion of equipment production data collection, enables the platform to achieve comprehensive digital management of CNC equipment by constructing a unified equipment digital model and ledger management system.

[0090] First, the platform establishes a structured data model for each device using standardized modeling methods. This model covers static attributes such as device number, type, control system, production line, supported processes, communication protocols, and interface parameters. It also defines the device's data acquisition capabilities, including the types of operating parameters, physical units, sampling frequency, and alarm thresholds. All device models have universal templates and personalized expansion capabilities, supporting flexible expansion for subsequent integration.

[0091] Based on the structured modeling of equipment, the system synchronously establishes an equipment ledger, assigns a unique digital identity ID to each piece of equipment, and records its entire lifecycle information, including procurement information, commissioning records, commissioning date, maintenance history, parts replacement records, and maintenance plans. At the same time, it is deeply integrated with modules such as real-time operation data, alarm records, and energy consumption analysis to form a complete chain from static description to dynamic monitoring.

[0092] Through equipment modeling, the platform can dynamically assess equipment status, calculate the Equipment Health Index (EHI), and support intelligent maintenance reminders based on runtime or status. Simultaneously, the system supports mobile scanning to view equipment records, quickly locate operational data, alarm history, and maintenance recommendations, improving fault response efficiency and asset management accuracy.

[0093] The data visualization module features a professional large-screen dashboard that collects data on the operating status of the application monitoring system, enabling remote viewing and control. The large screen supports real-time refresh, automatic slideshow, and interactive clicks, and is compatible with touchscreens or remote large-screen control. The dashboard is laid out according to function and data hierarchy, highlighting key equipment status information and anomaly warnings.

[0094] The large-screen dashboard displays the current production line and workshop name, system time, and currently logged-in user identity; the factory equipment topology map shows the location and status (running, stopped, alarm) of all equipment in a bird's-eye view of the workshop, with equipment icons supporting real-time flashing and color changes, and clicking to display details; core equipment operating parameters display indicators such as spindle speed, load, vibration, and temperature of the currently selected equipment, presented in a large dashboard and real-time curve graph; alarm distribution and risk heat map display information such as the total number of alarms, alarm level percentage, and distribution of affected equipment, with alarm information supporting real-time updates and severity indicated by color; and an abnormal event chain view displays the causal chain of typical alarms, combined with process parameters and historical data, to help managers trace the causes of anomalies.

Claims

1. A CNC equipment working data acquisition and processing platform, characterized in that, The platform is built with a modular architecture and includes five modules: data acquisition module, data storage module, data retrieval module, device modeling module, and data visualization module. Data acquisition module: It completes the acquisition of CNC equipment operation data through protocol plug-in, edge intelligent acquisition and anomaly recognition mechanism; Data storage module: Constructs a multi-layered industrial data storage system through edge caching, streaming processing, tiered cold and hot storage, and semantic asset management; Data access module: Provides standardized, multi-type data interfaces through a unified data service gateway to enable data access and system integration; Equipment modeling module: By building structured equipment models and complete lifecycle ledgers, it enables digital identity management of equipment, definition of operational capabilities, and assessment of health status, providing maintenance support; Data visualization module: Displays workshop topology, equipment status, operating parameters, alarm information and abnormal chain through professional large screen dashboards, enabling real-time interactive and intuitive monitoring of production.

2. The CNC equipment working data acquisition and processing platform according to claim 1, characterized in that, The data acquisition module includes: Various industrial communication protocols of different brands of equipment are encapsulated into independent protocol plug-in modules. These plug-ins are registered to the protocol manager through a unified interface standard. When a new protocol or device model is added, a new plug-in is introduced. The protocol plug-ins are hot-swappable. Protocol drivers and rule configurations can be added, updated, and uninstalled in real time during system operation. Devices can be dynamically connected to the platform as needed. At the forefront of data acquisition, an industrial-grade edge gateway device integrating a TPM chip is deployed. Each piece of collected data is generated within the edge gateway and then stamped with a timestamp, device ID, and digital signature. The platform deploys a lightweight blockchain, Hyperledger Fabric, which hashes the key raw data to form a data digest. The data digest is recorded on the blockchain, while the raw collected data is stored in a local or cloud database. An event-triggered hybrid sampling mechanism is adopted, which adds an event detection subsystem to the conventional timed polling. Event triggering conditions are predefined. When an event is detected, the high-frequency sampling mode is switched, and the past low-frequency data is windowed and re-sampled to collect complete data before and after the event. This data is analyzed and judged locally at the edge, prioritized, and key data is pushed to the central system in real time in sequence using the MQTT protocol. The edge acquisition layer nodes deploy intelligent acquisition guardian agents, which monitor the acquisition frequency, data fluctuation trends and acquisition stability in real time based on an anomaly detection model. They detect potential anomalies such as sensor failure, data interruption and buffer overflow, and issue warnings or automatically restart the acquisition service. For missing data segments, they use a pre-trained time series prediction model LSTM to estimate and complete them, and mark and distinguish the completed results to strictly separate them from the original acquisition data. We construct a semantic-driven alarm engine and a graph neural network anomaly detection mechanism to identify complex interaction patterns and weak signal anomaly combinations from multiple data sources, enabling early prediction of faults. We use a multi-channel alarm push system, including SMS, voice calls, App push, and email, to automatically categorize and push alarms based on the urgency and business impact level of self-built events.

3. The CNC equipment working data acquisition and processing platform according to claim 1, characterized in that, The data storage module includes: The underlying edge caching layer is deployed near the acquisition terminal to receive and cache the raw data uploaded by the acquisition device; The second real-time stream processing layer performs sequential control, source marking, and delivery of the data stream; The third layer, a cold and hot separation storage layer, separates and stores data. Hot data is stored in a high-performance time-series database or columnar database for recently accessed high-frequency data. Cold data is stored in a distributed file system in an archive and compressed manner for historical tracing, long-term analysis, and AI model training. The topmost semantic data asset layer stores alarm event chains, equipment operation graphs, and knowledge associations.

4. The CNC equipment working data acquisition and processing platform according to claim 1, characterized in that, The data retrieval module includes: A unified data service gateway is built as a system bus, which encapsulates various types of data interfaces, including real-time data interfaces, historical data interfaces, alarm event interfaces, and data export interfaces for AI training and BI analysis. All calls are authenticated, routed, and responded through this gateway, with controllable access permissions, unified interface calls, and auditability. The interface design adopts the standardized REST and WebSocket protocols.

5. The CNC equipment working data acquisition and processing platform according to claim 1, characterized in that, The equipment modeling module establishes a digital equipment model and ledger management system, including: Establish a structured data model for each device, covering device number, type, control system, production line, supported processes, communication protocol, interface parameter static attributes, define its data acquisition capability model, including the types of operating parameters, physical units, sampling frequency and alarm thresholds, and all device models have a general and extensible template; Establish an equipment ledger, assign a unique digital identity ID to each piece of equipment, and record its procurement information, commissioning records, production date, maintenance history, parts replacement records, and maintenance plans throughout its entire lifecycle. At the same time, integrate it with real-time operation data, alarm records, and energy consumption analysis modules to form a complete chain. By modeling the equipment, the platform dynamically assesses the equipment status, calculates the Equipment Health Index (EHI), and provides intelligent maintenance reminders.

6. The CNC equipment working data acquisition and processing platform according to claim 1, characterized in that, The data visualization module features a professionally designed large-screen dashboard, including: The large-screen dashboard displays the current production line and workshop name, system time, and the identity of the currently logged-in user; The factory equipment topology map displays the location and status of all equipment in a bird's-eye view of the workshop. Equipment icons support real-time flashing and color changing, and clicking on them will display details. The core operating parameters of the equipment are displayed, showing the spindle speed, load, vibration, and temperature of the currently selected equipment, presented in a large dashboard and real-time graph format. Alarm distribution and risk heat map displays the total number of alarms, the percentage of alarm levels, and the distribution of devices experiencing alarms. Alarm information can be refreshed in real time, and severity is indicated by color. The abnormal event chain view displays the causal chain of typical alarms, and combines process parameters with historical data to help managers trace the causes of abnormalities.

Images