Intelligent factory control system and control method based on distributed architecture

Abstract

The invention discloses an intelligent factory control system and method based on a distributed architecture. The control system comprises a production control unit and a data processing and storage unit, the production control unit controls production equipment of the intelligent factory; the data processing and storage unit comprises a data acquisition assembly, a data pre-aggregation assembly and a database, and the data pre-aggregation assembly is used for performing pre-aggregation processing on the original production data acquired by the data acquisition assembly to form structured production data and writing the structured production data into the database for storage. The control method is implemented by adopting the control system. According to the method, the data writing performance is remarkably improved, the real-time writing bottleneck is avoided, the efficient operation of the system is ensured, the reliability and stability of the production system are improved, the single-point fault risk is reduced, the safety of data assets and the stability and performance of a production core database are enhanced, and the intelligent level and the operation efficiency of a factory are improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing technology, and in particular relates to a control system and method for an intelligent factory. Background Technology

[0002] Traditional factory production control systems typically employ a centralized architecture, which suffers from high single-point-of-failure risk, limited data processing capabilities, low data storage efficiency, and poor data security. With the development of Industry 4.0 and smart manufacturing, higher demands are placed on the reliability, real-time performance, data processing capabilities, and data security of production systems. Especially for advanced manufacturing enterprises pursuing a "lights-out factory" model, ensuring stable production line operation, reliable and secure production data, and enabling production optimization through data analysis in an unattended environment is a pressing technical challenge. Existing technologies for data acquisition, storage, and backup often rely on direct writing to databases, which can easily create database write bottlenecks and impact system performance. Data backup is also mostly limited to local or same-datacenter backups, leaving the risk of data loss in the event of a major incident. Furthermore, traditional systems have limited analytical capabilities for production data, making it difficult to effectively support advanced applications such as predictive maintenance and energy optimization. Therefore, a novel technical solution is needed to address these issues. Summary of the Invention

[0003] The technical problem to be solved by this invention is to address the issues of low data storage efficiency, weak data security, and poor control stability in existing smart factories, and to propose a smart factory control system and control method based on a distributed architecture.

[0004] To solve the above-mentioned technical problems, the technical solution proposed by this invention is as follows: A smart factory control system based on a distributed architecture includes a production control unit and a data processing and storage unit. The production control unit interacts with the production line of the smart factory to control the production equipment of the smart factory. The data processing and storage unit includes a data acquisition component, a data pre-aggregation component, and a database. The data acquisition unit is used to acquire raw production data from the production control unit and the production equipment. The data pre-aggregation component is used to pre-aggregate the raw production data acquired by the data acquisition component to form structured production data, and then write it into the database for storage.

[0005] This invention incorporates a data pre-aggregation component. Compared to the conventional direct collection-storage method, the "aggregate first, then store" strategy of this invention significantly reduces the number of database write operations and the amount of data, thereby greatly improving database write performance and overall system throughput. Simultaneously, the distributed PLC cluster design significantly improves system redundancy and reliability, avoiding the single point of failure risk of traditional centralized control. When a PLC fails, other PLCs can continue production, or a backup PLC can take over its control tasks, ensuring continuous operation of the production line.

[0006] As a further preferred embodiment of the above technical solution, the database includes a main database and a query database. The query database is physically and / or logically separated from the main database and obtains structured production data from the main database. The query database is used to respond to data query and data analysis requests from the intelligent factory control system. The query database, being physically or logically separated from the main database, primarily serves to respond to various data query and analysis requests, such as report generation, historical data analysis, and visualization. By separating the query load from the write load, the security and performance of the main database can be guaranteed, and interference with core production data storage caused by complex queries can be avoided.

[0007] As a further preferred embodiment of the above technical solution, the intelligent factory control system also includes a data backup unit. This data backup unit is used to back up data stored in the database, and it is configured separately from the main database. This means that the data in the main database is not only backed up locally, but also transmitted via network to backup servers or storage systems located in other factories or even remote locations. This distributed backup mechanism can effectively mitigate the risk of data loss due to localized catastrophic physical accidents (such as fires, earthquakes, power outages, equipment damage, etc.). Even if the entire production plant suffers irreversible damage, core production data can still be recovered remotely, ensuring the security of the enterprise's data assets and business continuity.

[0008] Based on the same technical concept, this invention also provides a smart factory control method based on a distributed architecture, implemented using the aforementioned smart factory control system, comprising the following steps: Production control steps: Control the production equipment in the smart factory through the production control unit; Data acquisition steps: Collect raw production data through the data acquisition component; Data pre-aggregation step: The collected discrete raw production data is pre-aggregated to generate structured production data; Data storage step: Store the structured data in the database.

[0009] As a further preferred embodiment of the above technical solution, the multiple PLC clusters realize collaborative control, fault detection and handling, and load balancing functions through a distributed communication protocol. The fault detection and handling includes real-time monitoring of the heartbeat signal of the PLC unit and fault transfer control; collaborative control is achieved through state synchronization and task scheduling among multiple PLCs; and load balancing dynamically adjusts task allocation based on real-time load data.

[0010] As a further preferred embodiment of the above technical solution, the fault detection and processing function is implemented through the following method: In the production control step, the real-time heartbeat signals of all PLC units within each PLC cluster in the production control unit are detected. When an abnormal real-time heartbeat signal of a certain PLC unit is detected, it is determined whether there is data loss in that PLC unit. If there is data loss, the PLC cluster triggers fault isolation and takes over the control task after querying and synchronizing the latest data through another backup PLC unit. If there is no data loss, the control task is directly taken over through another backup PLC unit.

[0011] As a further preferred embodiment of the above technical solution, the pre-aggregation process is a process of integrating a series of discrete raw production data into a structured production event with business significance, including: (1) Based on the type, collection frequency and business scenario of the original production data, different time window parameters are set to limit the time range of pre-aggregation and form structured production events within the time window; (2) Based on the status parameters of production equipment and the priority of production tasks in the original production data, dynamically select the corresponding pre-aggregation strategy to form structured production events; the pre-aggregation strategy includes, but is not limited to: a) For continuous measurement data in the original production data, when the data fluctuation is small and the equipment is running stably, the arithmetic mean of the data within the time window is used for pre-aggregation to reflect the balanced operation status of the equipment. b) For event count data in the raw production data, during peak production periods, a counting strategy is used to count the frequency of events, which facilitates the assessment of equipment load and abnormal conditions. c) For cumulative data in the raw production data, summation is used to ensure the accurate accumulation of key information such as total output; d) For extreme value indicators in the raw production data, extract the maximum or minimum value within the time window to detect the upper and lower limits of equipment performance and abnormal fluctuations; (3) The pre-aggregation strategy is automatically adjusted based on the real-time production equipment status monitoring results and historical production data trends in the original production data to optimize the pre-aggregation accuracy and response speed; (4) For special business scenarios, the preset custom aggregation logic is triggered to form structured production events for special scenarios; flexible event definition is supported to avoid using a fixed exhaustive method to preset all structured events, thereby improving the system's adaptability and scalability.

[0012] As a further preferred embodiment of the above technical solution, the intelligent factory control method also includes a data backup step: periodically or in real time, transmitting the data of the database to one or more backup nodes that are physically isolated from the main database for storage.

[0013] As a further preferred option of the above technical solution, the intelligent factory control method also includes an intelligent analysis step: the structured production data that has been collected and stored in the database is vectorized, and the vectorized data is deeply analyzed by constructing and training a time series analysis model to achieve prediction and optimization of production status.

[0014] As a further preferred embodiment of the above technical solution, the prediction and optimization of production status includes the prediction of equipment status and / or the optimization of the production process.

[0015] As a further preferred embodiment of the above technical solution, the specific operations of the intelligent analysis step include: (1) For the structured production data stored in the database, according to the predetermined dimensions and features, the structured production data is converted into vector form through normalization, feature selection and time-series coding methods; (2) Construct an advanced time series analysis model based on Long Short-Term Memory Network (LSTM) or Variational Autoencoder (VAE), and train the model using labeled or unlabeled historical production data. The training process includes loss function design, optimization algorithm selection and training parameter adjustment. (3) Use the trained time series model to perform in-depth analysis on the latest vectorized data to predict future production status indicators (such as equipment failure probability, capacity fluctuation, etc.). (4) Based on the model prediction results, production parameters are automatically adjusted in conjunction with feedback control rules to achieve dynamic optimization of the production process. Based on deep time series learning and intelligent data processing technology, traditional statistical analysis is elevated to an intelligent and dynamically adaptive level. Efficient modeling of complex nonlinear time series relationships and utilization of diverse production data can better capture complex nonlinear time series relationships and dynamic changes in the production environment, improve prediction accuracy and optimization effect, realize higher precision state prediction and capacity optimization, and thus enhance the autonomous adjustment capability and operating efficiency of the smart factory, which is significantly better than traditional prediction and production methods.

[0016] Compared with the prior art, the advantages of the present invention are as follows: 1. This invention employs data pre-aggregation technology to integrate discrete data points in memory before writing them to the database, which significantly improves data writing performance, avoids real-time writing bottlenecks, and ensures efficient system operation.

[0017] 2. This invention implements local redundancy configuration for key equipment and control units in the production line, and combines real-time data synchronization and automatic fault switching of a distributed PLC cluster to construct a dual redundancy mechanism, thereby improving the reliability and stability of the production system and reducing the risk of single point of failure.

[0018] 3. This invention greatly enhances the security of data assets through a primary-backup database architecture and a distributed cross-plant backup solution, effectively preventing data loss even in the event of a localized catastrophic accident.

[0019] 4. This invention introduces a design that separates the query database from the main database, ensuring the stability and performance of the core production database while meeting the needs of big data analysis.

[0020] 5. By applying production data vectorization and time-series analysis, this invention enables the system to predict equipment failures and optimize production energy consumption, thereby improving the factory's intelligence level and operational efficiency.

[0021] The above advantages make this invention particularly suitable for lights-out factories, unmanned production lines, continuous production processes (such as metallurgy, chemical industry, power industry, etc.), and any intelligent manufacturing scenario with extremely high requirements for system reliability, data processing performance, and data security. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the overall architecture of the intelligent factory control system based on distributed architecture of the present invention.

[0024] Figure 2 This is a flowchart of the PLC cluster fault self-healing process in the production control steps of the intelligent factory control method based on distributed architecture of the present invention. Detailed Implementation

[0025] To facilitate understanding of the present invention, the present invention will be described more fully and in detail below with reference to the accompanying drawings and preferred embodiments, but the scope of protection of the present invention is not limited to the following specific embodiments.

[0026] It should be noted that when a component is described as being "fixed to, attached to, connected to or connected to" another component, it can be directly fixed to, attached to, connected to or connected to the other component, or it can be indirectly fixed to, attached to, connected to or connected to the other component through other intermediate connectors.

[0027] Unless otherwise defined, all technical terms used herein have the same meaning as commonly understood by those skilled in the art. The technical terms used herein are for the purpose of describing particular embodiments only and are not intended to limit the scope of the invention.

[0028] Unless otherwise specified, all raw materials, reagents, instruments and equipment used in this invention can be purchased from the market or prepared by existing methods.

[0029] Example 1: like Figure 1 As shown, the intelligent factory control system based on a distributed architecture in this embodiment includes a production control unit, a data processing and storage unit, and a data backup unit.

[0030] The production control unit interacts with the production line of the smart factory to control the production equipment. The production control unit includes multiple PLC clusters that communicate with sensors and actuators on the production line. The PLC units in these PLC clusters are connected to various sensors (such as temperature sensors, pressure sensors, flow sensors, vibration sensors, etc.) and actuators (such as valves, motors, heaters, etc.) on the thermal production line via industrial Ethernet or other fieldbuses. Each PLC cluster is responsible for controlling a specific area or a specific type of production equipment and collecting relevant production status data in real time. Multiple PLC clusters are distributed and deployed in different locations in the smart factory according to the area or type of production equipment they control.

[0031] The data processing and storage unit includes a data acquisition component, a data pre-aggregation component, and a database. The data acquisition unit, consisting of a data acquisition device equipped with acquisition software, is used to efficiently and in real-time collect raw production data from the PLC cluster and various sensors on the production equipment. The data pre-aggregation component pre-aggregates the raw production data collected by the data acquisition component to form structured production data before writing it into the database for storage. The pre-aggregation component utilizes dedicated equipment or industrial computing platforms with strong data processing capabilities, such as industrial dedicated servers, edge computing devices, dedicated FPGA-based accelerator cards, and high-performance embedded processing modules; the specific configuration depends on the production situation. Upon receiving the raw, discrete production status data transmitted by the acquisition software, the data pre-aggregation component does not immediately write the data to the database but instead briefly stores and preprocesses the data in system memory. The database includes a main database and a query database. The query database is physically and / or logically separated from the main database and obtains structured production data from the main database. The query database responds to data query and data analysis requests from the intelligent factory control system. The main database is the core data storage for production and is designed with high availability and stability in mind, avoiding direct exposure to high-concurrency real-time write pressure. Database queries can retrieve data from the master database using ETL (Extract-Transform-Load) tools or real-time data synchronization mechanisms. Its primary responsibility is to respond to various data query and analysis requests, such as report generation, historical data analysis, and visualization.

[0032] The data backup unit is used to back up the data stored in the database. The data backup unit and the main database are set up separately. The backup can be a combination of periodic full backup and incremental backup, or it can be a real-time streaming backup.

[0033] The intelligent factory control method based on a distributed architecture in this embodiment is implemented using the intelligent factory control system of this embodiment, and includes the following steps: Production control steps: Controlling the production equipment in the smart factory through the production control unit. In the production control steps, such as... Figure 2 As shown, each PLC cluster in the production control unit monitors the real-time heartbeat signals of all its internal PLC units. When an abnormal real-time heartbeat signal is detected in a PLC unit, it is determined whether data loss has occurred. If data loss is found, the PLC cluster triggers fault isolation and takes over the control task after querying and synchronizing the latest data through another backup PLC unit. If no data loss is found, the control task is directly taken over by another backup PLC unit. After taking over the control task, an alarm is sent to the central control room, and the status is updated to ensure uninterrupted production.

[0034] Data acquisition steps: Raw production data is collected through data acquisition components; various status data during the production process are continuously and in real time collected by a large number of distributed sensors deployed on the production line. This data includes equipment operating parameters (such as motor speed, pump pressure, valve opening), environmental parameters (such as temperature, humidity, gas concentration), and product quality parameters (such as size, weight, color). The collected data is transmitted to the data acquisition unit via fieldbus or industrial Ethernet for processing by the acquisition software.

[0035] The data pre-aggregation step involves pre-aggregating the collected discrete raw production data to generate structured production data. Before writing the data to permanent storage (i.e., the main database), the collected discrete raw production data is first sent to the data pre-aggregation component in system memory. In this component, efficient algorithms and memory buffers are used to perform real-time or near-real-time aggregation processing on the raw data. For example, hundreds of temperature readings collected per second can be aggregated into an average temperature per minute; a series of vibration sensor data can be aggregated into a single abnormal equipment vibration event; or multiple related parameters can be integrated into a structured production batch record. The core of this step is to transform "data points" into "meaningful production events" or "structured data blocks," thereby significantly reducing the frequency and volume of subsequent database writes.

[0036] Data tiered storage steps: Store structured data in a database.

[0037] Distributed data backup steps: To ensure the absolute security of data assets, data from the primary database is periodically or in real-time transferred to one or more backup nodes physically isolated from the primary database for storage. These backup nodes can be different servers located within the same facility, but more importantly, they include remote backup servers located in different facilities or geographical locations. This cross-facility / cross-regional distributed backup strategy effectively prevents permanent data loss due to localized physical accidents (such as fires, power outages, equipment damage, etc.), thereby ensuring business continuity and data disaster recovery capabilities.

[0038] Intelligent Analysis Steps: To achieve a higher level of factory intelligence, the production data collected and stored in the main database or query database is vectorized. For example, multiple sensor readings (such as vibration, temperature, current, and voltage) over a period of time are combined into a high-dimensional vector. Then, by building and training time-series analysis models (such as LSTM, GRU, Transformer, ARIMA, etc.), in-depth analysis is performed on the vectorized data, and further actions are taken, such as: 1. Equipment Status Prediction: By analyzing the time-series change patterns of equipment operating parameters, the wear level of equipment components, the probability of failure, and the remaining useful life (RUL) can be predicted, thereby guiding predictive maintenance, avoiding unplanned downtime, and extending equipment life. 2. Production Energy Consumption Optimization: By analyzing the time-series relationship between energy consumption data (such as electricity, water, steam, and gas) in the production process and production status and environmental conditions, abnormal energy consumption patterns can be identified, production process parameters can be optimized, energy conservation and emission reduction can be achieved, and operating costs can be reduced.

[0039] Taking production process optimization as an example, the intelligent analysis step can be carried out using the following methods (other existing conventional optimization methods can also be selected), including: (1) Data preprocessing and vectorization: For structured production data stored in the database, according to predetermined business dimensions and characteristics, conventional normalization methods (such as Min-Max normalization or Z-score standardization) are first used for numerical standardization to improve the stability of model training. Subsequently, key features are screened using feature selection methods (such as variance-based filtering, correlation analysis, or principal component analysis, etc.) to discard redundant and noisy data, thereby reducing model complexity. For time series data, the sliding window method is used to slice the time series data and perform time series encoding (such as timestamp encoding, periodic feature encoding) to generate fixed-length feature vectors, which can be described by the following formula: eigenvector x t =[ f 1( t ), f 2( t ),..., fn ( t ), τ ( t ),sin( T 2 πt ),cos( T 2 πt )]; in, fi ( t Let be the value of the i-th feature at time t. τ ( t The first two items represent timestamp features, and the last two items are periodic time codes.

[0040] (2) Model construction and training: A time series analysis model based on Long Short-Term Memory (LSTM) networks or Variational Autoencoders (VAEs) is adopted. The model structure design adopts multi-layer LSTM unit stacking or VAE encoder-decoder framework to achieve efficient modeling of time series data. During training, historical production data can include labeled data (such as data with fault labels) or unlabeled data (normal operation data). Labeling methods can be generated by manual diagnosis or automatic detection systems. The loss function design depends on the specific task: for example, cross-entropy loss function is used for fault prediction, and mean squared error (MSE) loss function is used for reconstruction tasks. The optimization algorithm generally adopts Adam or RMSProp to adapt to the training requirements of non-stationary time series data. The training parameters (learning rate, batch size, number of iterations, etc.) are adjusted through cross-validation and early stopping to prevent overfitting.

[0041] (3) Prediction and analysis: Using the trained time series model, forward inference is performed on the latest vectorized data to output predictions of production status indicators for a future period, such as equipment failure probability and capacity fluctuation trend. Anomaly warnings are assessed based on probability thresholds and prediction confidence.

[0042] (4) Dynamic optimization control: Based on the prediction results, production parameters are automatically adjusted according to the preset feedback control rules (such as conventional control methods such as PID control and fuzzy logic control) to achieve dynamic optimization of the production process and improve production efficiency and equipment reliability.

[0043] This intelligent analysis module typically operates through a separate data analysis platform or database query, and does not directly impact the performance of the main production database.

[0044] Taking a thermal production line as an example, sensor data such as furnace temperature, pressure, gas flow rate, and flue gas temperature are collected in real time by a distributed PLC cluster. This raw production data enters the data pre-aggregation component. Here, hundreds of temperature readings collected every second are aggregated into average and maximum temperatures per minute. Simultaneously, instantaneous fluctuations in gas flow rate are also integrated into gas consumption events. This aggregated structured data (e.g., "Furnace average temperature 681℃, maximum 691℃, October 26, 2024, 10:30-10:31") is written to the main database. Meanwhile, data from the main database is periodically (e.g., hourly) transmitted via a dedicated line to a backup database located 50 kilometers away in another plant. When managers need to view production data trends over the past year to optimize the operating parameters of the air-cooled and water-cooled systems, they perform complex time-series analysis queries on the database; these queries do not affect the normal write operations of the main database. Meanwhile, based on historical cooling data and equipment wear data, the intelligent analysis module predicts that the bearing of the cooling pump of the No. 15 die-casting machine cooling station on Line B of the thermal engineering phase 1 of the production line may fail within the next 30 days, and issues a maintenance warning in advance, thereby effectively avoiding sudden shutdowns.

[0045] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application. The above are merely preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. For those skilled in the art, improvements and modifications obtained without departing from the technical concept of the present invention should also be considered within the protection scope of the present invention.

Claims

1. A smart factory control system based on a distributed architecture, characterized in that, The system includes a production control unit and a data processing and storage unit. The production control unit interacts with the production line of the smart factory to control the production equipment. The production control unit includes multiple PLC clusters that communicate with sensors and actuators on the production line. These multiple PLC clusters are distributed and deployed in different locations within the smart factory according to the area or type of production equipment they control. The data processing and storage unit includes a data acquisition component, a data pre-aggregation component, and a database. The data acquisition unit is used to acquire raw production data from the production control unit and the production equipment. The data pre-aggregation component is used to pre-aggregate the raw production data acquired by the data acquisition component to form structured production data, and then write it into the database for storage.

2. The intelligent factory control system based on a distributed architecture according to claim 1, characterized in that, The database includes a main database and a query database; the query database is physically and / or logically separated from the main database and obtains structured production data from the main database. The query database is used to respond to data query and data analysis requests from the intelligent factory control system.

3. The intelligent factory control system based on a distributed architecture according to claim 1 or 2, characterized in that, The intelligent factory control system also includes a data backup unit, which is used to back up the data stored in the database, and the data backup unit and the main database are arranged separately.

4. A smart factory control method based on a distributed architecture, characterized in that, The intelligent factory control system according to any one of claims 1-3 is used to implement the system, comprising the following steps: Production control steps: Control the production equipment in the smart factory through the production control unit; Data acquisition steps: Collect raw production data through the data acquisition component; Data pre-aggregation step: The collected discrete raw production data is pre-aggregated to generate structured production data; Data storage step: Store the structured data in the database.

5. The intelligent factory control method based on a distributed architecture according to claim 4, characterized in that, The multiple PLC clusters implement fault detection and handling functions through a distributed communication protocol.

6. The intelligent factory control method based on a distributed architecture according to claim 5, characterized in that, The fault detection and handling function is implemented through the following method: In the production control step, the real-time heartbeat signal of all PLC units in each PLC cluster of the production control unit is detected. When an abnormal real-time heartbeat signal of a certain PLC unit is detected, it is determined whether there is data loss in that PLC unit. If there is data loss, the PLC cluster triggers fault isolation and takes over the control task after querying and synchronizing the latest data through another backup PLC unit. If there is no data loss, the control task is taken over directly through another backup PLC unit.

7. The intelligent factory control method based on a distributed architecture according to claim 4, characterized in that, The pre-aggregation process is the process of integrating a series of discrete raw production data into structured production events with business significance, including: Based on the type, collection frequency, and business scenario of the original production data, different time window parameters are set to limit the time range of pre-aggregation and form structured production events within the time window; And / or, based on the status parameters of production equipment and the priority of production tasks in the original production data, the corresponding pre-aggregation strategy is dynamically selected to form structured production events; during the process, the pre-aggregation strategy is automatically adjusted based on the real-time production equipment status monitoring results and historical production data trends in the original production data. And / or for specific business scenarios, trigger preset custom aggregation logic to form structured production events for specific scenarios.

8. The intelligent factory control method based on a distributed architecture according to claim 4, characterized in that, It also includes a data backup step: periodically or in real time, transferring the data of the database to one or more backup nodes that are physically isolated from the main database for storage.

9. The intelligent factory control method based on a distributed architecture according to claim 4, characterized in that, It also includes intelligent analysis steps: the structured production data that has been collected and stored in the database is vectorized, and the vectorized data is analyzed in depth by building and training a time series analysis model to achieve prediction and optimization of production status.

10. The intelligent factory control method based on a distributed architecture according to claim 9, characterized in that, Predicting and optimizing production status includes predicting the status of production equipment and / or optimizing the production process.

Images