An automation control platform operation data management method

Abstract

The application relates to the technical field of data processing, and discloses an automatic control platform operation data management method. The method comprises the following steps: collecting industrial equipment data through a field bus, storing the data in a database in a classified manner and establishing an index table, extracting a control loop to construct a DCS model, generating a process abnormality diagnosis report, calculating control parameters and outputting control instructions by using a self-tuning algorithm, optimizing a process formula based on implementation effects and constructing a recommended knowledge base. The application realizes an intelligent data management method for accurately diagnosing process abnormalities, automatically optimizing control parameters and continuously accumulating process knowledge, so that the control precision, diagnosis efficiency, parameter optimization capability and knowledge reuse level of an industrial production process are improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method for managing operational data of an automated control platform. Background Technology

[0002] Industrial automation control systems are a core component of modern industrial production. Traditional automation control platforms primarily rely on distributed control systems (DCS) and programmable logic controllers (PLCs) to achieve basic data acquisition and monitoring functions. These systems connect various sensors and actuators via communication networks such as fieldbus and industrial Ethernet to collect process parameters such as temperature, pressure, and flow rate, and perform simple logic control and closed-loop regulation according to preset programs. In terms of data management, traditional methods mainly employ a fixed-period sampling and storage mechanism, recording the collected data into a relational database at uniform time intervals. Abnormal states are monitored through a simple threshold-based alarm mechanism, and basic trend charts and reporting functions are provided. Control parameter adjustment mainly relies on experienced operators manually setting parameters based on equipment operating conditions and process requirements. Classic PID tuning methods such as the Ziegler-Nichols method are commonly used to determine control parameters. Process recipe management uses a preset recipe library, selecting the corresponding recipe according to different product requirements.

[0003] However, as industrial production moves towards intelligence and precision, the data management methods of traditional automated control platforms have significant shortcomings. First, fixed-period sampling mechanisms cannot simultaneously optimize system resources and capture critical data, leading to the storage of large amounts of redundant data or the loss of important transient information. Second, simple data storage methods lack differentiated processing strategies for different types of signals, resulting in low storage efficiency and difficulty in efficient retrieval. Third, alarm mechanisms based on single-point thresholds fail to consider the correlation between parameters and the overall system state, generating numerous false alarms and missed alarms. Fourth, control model construction lacks a comprehensive analysis of the system's dynamic characteristics and parameter correlations, resulting in poor control accuracy and stability. Fifth, the PID parameter tuning process relies heavily on experience, lacking systematic and automated parameter optimization methods, making it difficult to adapt to changes in operating conditions. Finally, process formulation management lacks a closed-loop feedback and continuous optimization mechanism with control effects, hindering knowledge accumulation and experience transfer, making continuous improvement and optimization of the production process difficult. Summary of the Invention

[0004] This application provides an automated control platform operation data management method, which is an intelligent data management method for achieving accurate diagnosis of process anomalies, automatic optimization of control parameters, and continuous accumulation of process knowledge, thereby improving the control accuracy, diagnostic efficiency, parameter optimization capability, and knowledge reuse level of industrial production processes.

[0005] Firstly, this application provides a method for managing operational data of an automated control platform. This method includes: collecting the operating status and process parameters of industrial equipment from an industrial controller via a fieldbus network; generating industrial data frames containing control point information after edge processing by a PLC; classifying the industrial data frames according to their signal type (discrete, analog, and computational) and storing them in a real-time database and a historical database to establish a process parameter index table; extracting control loop data from the process parameter index table to construct signal flow and control dependencies between devices, forming a DCS control model; generating a process anomaly diagnosis report based on the DCS control model, combined with process parameter change trends and equipment operating status, using multi-dimensional alarm analysis technology; calculating optimal control parameters using a PID parameter self-tuning algorithm based on the process anomaly diagnosis report, while simultaneously verifying operational safety boundaries, and outputting actuator control commands; performing deviation analysis between the implementation effect of the actuator control commands and production targets, dynamically optimizing the process formula library, and constructing an intelligent production recommendation knowledge base.

[0006] Secondly, this application provides an automated control platform operation data management system, which includes:

[0007] The acquisition module is used to acquire the operating status and process parameters of industrial equipment with the industrial controller through the fieldbus network, and generate industrial data frames containing control point location information after edge processing by the PLC.

[0008] The storage module is used to classify and store the industrial data frames into the real-time database and the historical database according to the signal type of discrete quantities, analog quantities and computational quantities, and to establish a process parameter index table.

[0009] The module is used to extract control loop data from the process parameter index table, construct the signal flow direction and control dependency relationship between equipment, and form a DCS control model;

[0010] The generation module is used to generate a process anomaly diagnosis report based on the DCS control model, combined with the process parameter change trend and equipment operating status, through multi-dimensional alarm analysis technology.

[0011] The calculation module is used to calculate the optimal control parameters based on the process anomaly diagnosis report using a PID parameter self-tuning algorithm, while checking the operation safety boundary and outputting actuator control commands.

[0012] The analysis module is used to analyze the deviation between the implementation effect of the actuator control commands and the production target, dynamically optimize the process formula library, and build an intelligent production recommendation knowledge base.

[0013] A third aspect of the present invention provides a computer device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the computer device to execute the above-described automated control platform operation data management method.

[0014] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the above-described automated control platform operation data management method.

[0015] The technical solution provided in this application collects the operating status and process parameters of industrial equipment through a fieldbus network and industrial controller, and generates industrial data frames containing control point information after edge processing by a PLC. This achieves unified data acquisition and standardized processing of distributed industrial equipment, improving the integrity and accuracy of data acquisition. Based on the signal type of the industrial data frames, they are classified into discrete, analog, and computational quantities and stored in real-time and historical databases, respectively, and a process parameter index table is established. A differentiated data storage strategy is adopted, significantly optimizing storage efficiency and improving data retrieval performance. Control loop data is extracted from the process parameter index table to construct the signal flow and control dependency relationships between devices, forming a DCS control model. This enables the system to fully understand the topology and data flow paths of the industrial control system, providing a system-level perspective for subsequent analysis. Based on the DCS control model, combined with process parameter change trends and equipment operating status, a process anomaly diagnosis report is generated through multi-dimensional alarm analysis technology. This process applies artificial intelligence algorithms for multi-dimensional data fusion and correlation analysis, ensuring that the diagnosis results are not limited to single-point threshold judgments but are analyzed from the perspective of overall system behavior, significantly improving the accuracy of anomaly diagnosis. The accuracy of detection and the precision of root cause analysis, especially the application of algorithms such as Bayesian causal inference and principal component anomaly measurement, enable the system to accurately identify the root causes of anomalies and trace their propagation paths in complex industrial environments. Based on the process anomaly diagnosis report, the system uses a PID parameter self-tuning algorithm to calculate the optimal control parameters, while simultaneously verifying the operational safety boundary and outputting actuator control commands. In this step, the combined application of the improved Ziegler-Nichols algorithm and particle swarm optimization algorithm fully leverages the respective advantages of traditional control theory and modern optimization algorithms, achieving automatic optimization of control parameters and ensuring safety, significantly reducing the workload of manual parameter tuning and improving control accuracy. The system then performs deviation analysis between the implementation effect of actuator control commands and production targets, dynamically optimizes the process formula library, and constructs an intelligent production recommendation knowledge base. This step, through adaptive incremental learning algorithms and knowledge graph technology, establishes a complete knowledge accumulation and experience inheritance mechanism. The system can continuously learn from production practice and optimize process parameters, forming a virtuous cycle of continuous improvement. This achieves the structured expression and efficient reuse of industrial knowledge, providing data-driven intelligent recommendation support for decision-making under similar production conditions. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments 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.

[0017] Figure 1This is a schematic diagram of one embodiment of the automated control platform operation data management method in this application.

[0018] Figure 2 This is a schematic diagram of one embodiment of the automated control platform operation data management system in this application.

[0019] Figure 3 This is a schematic block diagram of the structure of the computer device in an embodiment of the present invention. Detailed Implementation

[0020] This application provides a method for managing operational data of an automated control platform. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0021] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the automated control platform operation data management method in this application includes:

[0022] Step S101: Collect the operating status and process parameters of industrial equipment with the industrial controller through the fieldbus network, and generate an industrial data frame containing control point information after edge processing by the PLC.

[0023] Step S102: Based on the signal type of the industrial data frame, classify and store it into the real-time database and historical database according to discrete quantity, analog quantity and computational quantity, and establish a process parameter index table.

[0024] Step S103: Extract control loop data from the process parameter index table, construct the signal flow direction and control dependency relationship between equipment, and form a DCS control model;

[0025] Step S104: Based on the DCS control model, combined with the process parameter change trend and equipment operating status, generate a process anomaly diagnosis report through multi-dimensional alarm analysis technology;

[0026] Step S105: Based on the process anomaly diagnosis report, use the PID parameter self-tuning algorithm to calculate the optimal control parameters, and at the same time check the operation safety boundary, output the actuator control command;

[0027] Step S106: Analyze the deviation between the implementation effect of the actuator control command and the production target, dynamically optimize the process formula library, and build an intelligent production recommendation knowledge base.

[0028] It is understood that the executing entity of this application can be an automated control platform running a data management system, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiments use a server as an example for illustration.

[0029] Specifically, the operating status and process parameters of industrial equipment are collected via fieldbus networks and industrial controllers. Fieldbus networks are digital communication systems used to connect various devices in industrial environments, including standard protocols such as Profibus, Modbus, and Foundation Fieldbus, enabling stable data transmission even in harsh industrial environments. Industrial controllers refer to core industrial automation devices such as programmable logic controllers (PLCs) and distributed control systems (DCS) controllers, responsible for executing control logic and interacting with field devices. The collected data undergoes edge processing by the PLC. This edge processing refers to preliminary data filtering, verification, and compression operations performed at the data source to reduce invalid data transmission and improve data quality. The resulting industrial data frame is a structured data packet containing fields such as timestamp, device ID, parameter values, quality codes, and control point information. Control point information refers to the address identifier in the industrial control system corresponding to each data point, such as the I / O address or memory address in the PLC, or the point number in the DCS system. Categorizing and storing industrial data frames according to their signal types is a crucial step in effectively managing massive amounts of industrial data. Discrete quantities refer to signals with only a finite number of states, such as switch quantities and state quantities, like valve on / off states and motor operating states. Analog quantities refer to continuously changing physical quantity measurements, such as temperature, pressure, and flow rate. Computational quantities refer to secondary data calculated from other parameters, such as efficiency, output, and energy consumption. Real-time databases are specifically used to store the latest data on the current operating status, employing in-memory database technology, featuring high-speed read / write and low-latency characteristics, suitable for real-time monitoring and control. Historical databases are used for long-term storage of historical operating data, typically employing time-series database technology, supporting large-capacity storage and efficient time-series queries. During data storage, a process parameter index table is simultaneously established. This is a metadata structure that records information such as the name, ID, data type, unit, source device, storage location, and refresh frequency of each parameter, providing rapid location capabilities for subsequent data queries and analysis. Extracting control loop data from the process parameter index table is the foundation for constructing a system-level control model. A control loop refers to a closed-loop control unit formed from sensors, controllers, and actuators, such as a temperature PID control loop or a flow proportional control loop. By analyzing the data flow and signal transmission relationships between different control loops, signal flow diagrams and control dependency diagrams between devices can be constructed. Signal flow describes the path of data transmission from one device to another, while control dependencies reflect the master-slave relationship and control priority between devices. These relationships are ultimately summarized to form a DCS control model, a digital twin model describing the control logic architecture of the entire production system. It includes the topology of all control loops, device nodes, signal flows, and control dependencies, providing a system-level perspective for subsequent anomaly diagnosis and optimization.

[0030] Process anomaly diagnosis based on the DCS control model is a core function of the automation control platform. Multidimensional alarm analysis technology is not a simple single-point threshold judgment, but a complex analysis method that comprehensively considers multiple relevant parameters, historical trends, and the control model. It includes three key technologies: causal analysis based on the control model, time-series pattern recognition, and multi-parameter correlation analysis. Causal analysis based on the control model traces the upstream and downstream impact sources of abnormal parameters through the DCS control model, identifying root cause anomalies and secondary anomalies. Time-series pattern recognition technology identifies known types of anomalies by comparing parameter change patterns in a historical anomaly case library. Multi-parameter correlation analysis discovers abnormal correlations between parameters by calculating the time-series correlation and partial correlation coefficients. Through these analyses, the generated process anomaly diagnosis report includes anomaly type, root cause analysis, impact range, severity, and handling recommendations, providing a basis for process adjustment decisions. Optimizing control parameters based on the process anomaly diagnosis report is the execution phase of the automation control platform. The PID parameter self-tuning algorithm is a method for automatically adjusting proportional (P), integral (I), and derivative (D) parameters, finding parameter values ​​that enable the control system to be stable, fast, and accurate based on the system response characteristics. Commonly used PID self-tuning methods include relay feedback, model identification, and iterative learning. In this method, based on the problem type identified in the process anomaly diagnosis report, a suitable self-tuning algorithm is selected to calculate the optimal PID parameter values. Simultaneously, the calculated control parameters are checked for operational safety boundaries to ensure they remain within the equipment's permissible operating range, preventing equipment damage or safety accidents caused by extreme parameters. The final generated actuator control command includes new control parameter setpoints, execution time, and control strategy, and is distributed to the corresponding actuators via the control network.

[0031] Deviation analysis of the implementation effect of actuator control commands against production targets is a key step in continuous optimization. Deviation analysis compares the gap between actual production indicators (such as output, quality, energy consumption, etc.) and production targets, quantitatively assessing the effectiveness of adjustment measures. Based on the deviation analysis results, the process formula library is dynamically optimized. The process formula library is a database storing sets of process parameters for different products or batches, including product specifications, process parameter setpoints, control strategies, and other information. During dynamic optimization, the parameter settings in the process formula are updated and refined based on the deviation analysis results under different production conditions, gradually forming a parameter optimization model adapted to different production conditions. This optimization experience and knowledge are structured and stored to construct an intelligent production recommendation knowledge base. This knowledge base not only includes parameter optimization experience but also anomaly handling solutions, quality control points, and energy-saving and emission-reduction measures, providing intelligent decision support for subsequent production.

[0032] Taking a chemical production line as an example, this production line includes major equipment such as reactors, heat exchangers, separation towers, and storage tanks, which are connected to multiple Siemens S7-400 series PLC controllers via a Profibus-DP fieldbus network. In practical applications, process parameters such as temperature, pressure, flow rate, and liquid level of each device, as well as equipment status information such as valve status and pump operating status, are collected from the PLC and DCS systems. After edge processing by the PLC, these raw data are used to generate industrial data frames containing control point information. For example, the data frame of the reactor temperature sensor TIC-101 contains a temperature value of "135.6℃", a timestamp of "2023-05-15 10:30:15.023", a quality code of "GOOD", and a PLC data point address of "DB10.DBW20". These data frames are stored according to signal type. Analog quantities such as temperature and pressure are stored in the "ANALOG_TAG" table of the historical database, while discrete quantities such as valve status are stored in the "DISCRETE_TAG" table. Simultaneously, a process parameter index table containing over 5000 parameters is established. Control loop data is extracted from the index table to identify the control relationships between the temperature sensor TIC-101, the PID controller PIC-101, and the regulating valve TV-101 in the reactor temperature control loop, as well as their process connections with the upstream raw material preheater and downstream separation tower. A DCS control model containing 30 main control loops across the entire plant is then constructed. When the system detects increased reactor temperature fluctuations, multi-dimensional alarm analysis technology traces the issue through the DCS control model and identifies unstable upstream steam pressure as the root cause, generating an anomaly diagnostic report that includes root cause analysis and impact assessment. Based on the report, the PID parameter self-tuning algorithm calculates new control parameters for the steam regulating valve (adjusting the proportional band from 120% to 150% and extending the integral time from 60 seconds to 90 seconds), which are then sent to the control system after safety boundary verification. After 24 hours of execution, deviation analysis showed that the product qualification rate increased by 3 percentage points and energy consumption decreased by 2%. Based on this, the temperature control parameters of high viscosity product categories in the reaction process formula library were updated, and this optimization experience was added to the intelligent production recommendation knowledge base to provide a reference for similar working conditions.

[0033] In this embodiment, the operating status and process parameters of industrial equipment are collected via a fieldbus network and an industrial controller. After edge processing by a PLC, industrial data frames containing control point information are generated, achieving unified data acquisition and standardized processing of distributed industrial equipment, thus improving the integrity and accuracy of data acquisition. Based on the signal type of the industrial data frames, they are categorized into discrete, analog, and computational quantities and stored in real-time and historical databases, respectively, and a process parameter index table is established. A differentiated data storage strategy is adopted, significantly optimizing storage efficiency and improving data retrieval performance. Control loop data is extracted from the process parameter index table to construct the signal flow and control dependency relationships between devices, forming a DCS control model. This enables the system to comprehensively understand the topology and data flow paths of the industrial control system, providing a system-level perspective for subsequent analysis. Based on the DCS control model, combined with process parameter change trends and equipment operating status, a process anomaly diagnosis report is generated using multi-dimensional alarm analysis technology. This process applies artificial intelligence algorithms for multi-dimensional data fusion and correlation analysis, ensuring that the diagnostic results are no longer limited to single-point threshold judgments but are analyzed from the perspective of overall system behavior, significantly improving anomaly detection. The accuracy of measurements and the precision of root cause analysis, especially the application of algorithms such as Bayesian causal inference and principal component anomaly measurement, enable the system to accurately identify the root causes of anomalies and trace their propagation paths in complex industrial environments. Based on the process anomaly diagnosis report, the system uses a PID parameter self-tuning algorithm to calculate the optimal control parameters, while simultaneously verifying the operational safety boundary and outputting actuator control commands. In this step, the combined application of the improved Ziegler-Nichols algorithm and particle swarm optimization algorithm fully leverages the respective advantages of traditional control theory and modern optimization algorithms, achieving automatic optimization of control parameters and ensuring safety, significantly reducing the workload of manual parameter tuning and improving control accuracy. The system then performs deviation analysis between the implementation effect of actuator control commands and production targets, dynamically optimizes the process formula library, and constructs an intelligent production recommendation knowledge base. This step, through adaptive incremental learning algorithms and knowledge graph technology, establishes a complete knowledge accumulation and experience inheritance mechanism. The system can continuously learn from production practice and optimize process parameters, forming a virtuous cycle of continuous improvement. This achieves the structured expression and efficient reuse of industrial knowledge, providing data-driven intelligent recommendation support for decision-making under similar production conditions.

[0034] In one specific embodiment, the process of performing step S101 may specifically include the following steps:

[0035] (1) Use the distributed fieldbus interface to connect the communication ports of various industrial equipment to collect raw process parameters including temperature, pressure, flow rate, position and switch status;

[0036] (2) Select key control point data from the original process parameters, and use the sampling frequency adaptive adjustment mechanism to sample high-frequency changing parameters at the millisecond level and low-frequency changing parameters at the minute level;

[0037] (3) The sampled data is introduced into the PLC edge computing unit, and the data of different formats is unified into the target standard format through the data standardization conversion protocol;

[0038] (4) Perform signal and noise filtering on the target standard format data and use the sliding window midpoint filtering algorithm to remove abnormal fluctuation values;

[0039] (5) Add timestamps and device identifiers to the filtered data, and construct the metadata header information area and data payload area;

[0040] (6) Package and integrate the metadata header information area and the data payload area into an industrial data frame containing control point location information, and protect the data with a lightweight encryption protocol.

[0041] Specifically, a distributed fieldbus interface is used to connect the communication ports of various industrial devices to collect raw process parameters. The distributed fieldbus interface is the physical and protocol implementation of industrial communication networks, including standards such as Profibus, Modbus, and DeviceNet, and connects to different devices via RS-485 serial ports, Ethernet RJ45 interfaces, or HART protocols. The collected raw process parameters include various parameters such as temperature (-200℃ to 1600℃, measured by thermocouples and RTDs), pressure (0-20MPa, measured by pressure transmitters), flow rate (measured by electromagnetic flowmeters), position (measured by displacement sensors), and on / off status (Boolean data). These parameters directly reflect the operating status of the industrial equipment and the characteristics of the process. Screening key control point data from the collected raw parameters is a crucial step in optimizing data acquisition. Key control points refer to measurement points that have a significant impact on process control, such as reaction temperature control points and key quality parameters. The adaptive sampling frequency adjustment mechanism dynamically adjusts the sampling frequency based on the parameter change characteristics. High-frequency parameters such as motor speed are sampled at millisecond levels (10-100Hz), medium-frequency parameters such as temperature and pressure at second levels (0.1-1Hz), and low-frequency parameters such as material level at minute levels (0.01-0.1Hz). This mechanism monitors the parameter change rate; when the change rate exceeds a threshold, the sampling frequency is increased; when the parameter stabilizes, the sampling frequency is decreased, thus achieving rational resource allocation.

[0042] Sampling data is introduced into the PLC edge computing unit, a small computing device deployed at the data source, typically integrated within the PLC controller or existing as an expansion module. The PLC edge computing unit processes heterogeneous data through a data standardization conversion protocol, unifying different formats such as Modbus register data and OPC tag data into target standard formats like JSON or XML. The conversion process includes format unification (e.g., floating-point representation), unit conversion (imperial to metric), and encoding standardization (ASCII to Unicode), ensuring consistency and comparability of data from different sources. Noise filtering of the standardized data is crucial for improving data quality. Electromagnetic interference and sensor fluctuations in industrial environments generate noise, affecting control decisions. The sliding window median filtering algorithm sets a fixed-length window (usually 3-15 points) on the time series, sorts the data within the window, and selects the median value as the output; the window slides continuously as new data arrives. Compared to mean filtering, median filtering has better suppression of spike noise while preserving signal edge characteristics, making it suitable for industrial data processing. The window size selection must balance filtering effect and latency; a large window provides strong filtering but also increases latency, while a small window has the opposite effect.

[0043] The filtered data needs to be timestamped and device identifiers added to construct a metadata structure. The timestamp represents the UTC time information at the moment of acquisition, with millisecond-level accuracy; the device identifier follows the format "region-device type-serial number," uniquely identifying the data source. The metadata header contains descriptive information such as the timestamp, device ID, data type, quality code, and engineering units, while the data payload contains the actual numerical values. This hierarchical structure enables the receiving system to correctly parse and process the data. The metadata header and data payload are packaged into industrial data frames, which are data packets with a specific structure, including a frame header, metadata area, data area, and frame trailer. Control point information refers to the address of the data in the control system, such as the PLC's database address or variable name. The data frames are protected by lightweight encryption protocols such as AES-128. These algorithms ensure security while having low resource requirements, making them suitable for edge devices.

[0044] Taking the automation control of a paper mill as an example, the system connects the temperature transmitter (0-180℃), steam pressure transmitter (0-0.8MPa), and electric valves in the drying section via Profibus-DP. The drying temperature is selected as the key control point, and because its rate of change is approximately 0.5℃ / minute, it is set to sample every 5 seconds; the steam pressure fluctuates more rapidly, so it is set to sample every 200 milliseconds. The sampled data is processed by a Siemens S7-300 PLC edge unit, converting the Profibus messages into a unified JSON format. When the drying temperature exhibits a sequence of "125.2, 125.5, 131.8, 125.3, 125.6", a median filter with a window size of 5 outputs "125.5", successfully filtering out the outlier value of 131.8. After processing, the data is stamped with a timestamp of "2023-06-15T10:30:45.125Z" and the device ID "DRYER-SECTION5-TT103". Together with the value "125.5℃", it forms a standard data frame. After adding the control point information "DB15.DBD24", the data is transmitted with AES encryption to ensure data security and integrity.

[0045] In one specific embodiment, the process of performing step S102 may specifically include the following steps:

[0046] (1) Perform signal type analysis on industrial data frames and classify the input signals into discrete signals of switching quantity, analog signals of continuous change, and calculation signals of logic operation through signal feature recognition algorithm;

[0047] (2) For discrete signals of switching quantity type, a state change trigger recording mechanism is adopted to store the state transition point data into the discrete quantity data area of ​​the real-time database, while retaining the state duration information;

[0048] (3) For continuously changing analog signals, the data compression engine performs dead-zone compression on the original signal and writes the data points into the analog data area of ​​the real-time database only when the signal change exceeds the set threshold.

[0049] (4) For logic operation type calculation signals, the input source and calculation formula are recorded by the calculation dependency tracker, and the calculation result and formula are stored together in the calculation data area of ​​the real-time database.

[0050] (5) Based on the data lifecycle strategy, the time-rotation archiver periodically transfers historical data from the real-time database to the historical database and organizes the data structure according to the time partitioning storage method;

[0051] (6) Establish access path mapping for all process parameters in the real-time database and the historical database through the multidimensional index builder, and generate a process parameter index table containing parameter ID, data type, storage location and associated control loop information.

[0052] Specifically, upon receiving an industrial data frame, signal type analysis is performed. Signal type analysis is the process of classifying and judging data characteristics, using a signal feature recognition algorithm to categorize the input signal into three main types. This algorithm analyzes the data's value range, variation characteristics, and data source. The specific judgment logic is as follows: if the data has only a finite number of fixed values ​​(usually 0 / 1 or on / off) and the state transitions are significant, it is identified as a discrete signal of the switching quantity type; if the data exhibits a continuous numerical distribution and changes smoothly over time, it is identified as a continuously changing analog signal; if the data is a derived value generated from other data through calculation, it is identified as a computational signal of the logical operation type. The algorithm also analyzes metadata markers in the data frame, such as data type identifier bits, engineering units, and quality codes, to assist in determining the signal type. For data identified as discrete signals of the switching quantity type, a state change-triggered recording mechanism is used. This mechanism does not store data at fixed intervals but only records data when the signal state changes, significantly reducing storage requirements. The state change-triggered recording mechanism maintains a current state buffer, continuously comparing the differences between newly arrived data and the current state. When a state change is detected (e.g., from 0 to 1, from on to off), a record containing four key pieces of information is generated: the state value before the change, the state value after the change, the timestamp of the change, and the state duration. The state duration is obtained by calculating the difference between the current timestamp and the last state change timestamp, reflecting the continuous operating time of the device in a specific state, which is of great value for device operation condition analysis and fault diagnosis. These records are stored in the discrete data area of ​​the real-time database, which uses a data structure optimized for high-frequency writes and low-frequency queries, such as a log structure merged tree (LSM Tree).

[0053] For continuously changing analog signals, a data compression engine is used for processing. This is a storage optimization mechanism designed specifically for the characteristics of industrial time-series data. The data compression engine implements a dead-zone compression algorithm, which sets a change threshold (dead zone). A new data point is only written to the database when the absolute value of the difference between the new data point and the previous stored point exceeds this threshold. The dead zone size is usually set according to the engineering significance and accuracy requirements of the signal; for example, a dead zone of 0.5℃ might be set for temperature signals, and 0.01MPa for pressure signals. This method preserves the important trend of the signal's change while significantly reducing storage requirements. For slowly changing signals, a maximum time interval is set to ensure that sampling points are not missed for too long, typically 3-5 times the normal signal period. The compressed data points are written to the analog data area of ​​the real-time database. This area is optimized for time-series data characteristics and uses a time-series data structure for storage. For logical operation-type computational signals, a computation dependency tracker is used for processing. The computation dependency tracker is a mechanism for managing the lineage of derived data; it records the list of input source parameter IDs and the calculation formula for the computational signal. The input sources may be raw physical quantities or other computational quantities, forming a multi-layered data dependency network. Calculation formulas are stored as strings, containing mathematical expressions, function calls, and conditional logic. Whenever a computational signal is updated, the tracker simultaneously records the calculation result and the formula used for the calculation, storing both in the computational data area of ​​the real-time database. This design allows for not only traceability of the results but also understanding of the calculation methods, facilitating subsequent verification and auditing. Furthermore, when input source parameters are updated, the tracker automatically triggers the recalculation of relevant computational signals, ensuring data consistency.

[0054] As runtime progresses, historical data accumulated in the real-time database requires archiving. Based on the data lifecycle strategy, the time-rotating archiver periodically performs data migration tasks. The data lifecycle strategy defines data retention strategies for different stages based on data value, query frequency, and storage cost, typically including three levels: hot data (recent days, high-frequency access), warm data (recent weeks or months, medium-frequency access), and cold data (earlier history, low-frequency access). The time-rotating archiver automatically migrates data reaching a specific age from the real-time database to the historical database according to preset time rules (e.g., 2 AM daily), while performing necessary aggregation calculations, such as calculating hourly averages, maximums, and minimums. The historical database organizes data using a time-partitioned storage method, dividing data into different partitions according to time ranges (e.g., days, weeks, months). Each partition is an independent physical storage unit. This structure significantly improves the efficiency of queries by time range and facilitates the deletion of expired data. After data storage is completed, an efficient data access structure is created using a multi-dimensional index builder. The multidimensional index builder analyzes data characteristics and creates various index types, including hash indexes based on parameter IDs (for fast point queries), B+ tree indexes based on timestamps (for range queries), and association indexes based on control loops (for association analysis). These indexes collectively form an access path mapping between the real-time and historical databases, ultimately generating a process parameter index table. This index table is a metadata structure containing fields such as parameter ID, name, description, data type, engineering unit, acquisition frequency, storage location (database name, table name, partition information), data retention strategy, and associated control loop information. As the entry point for data access, the process parameter index table supports multi-dimensional parameter searches by name, type, control loop, and other dimensions, forming the foundation for subsequent data analysis and control optimization.

[0055] Taking the automated control platform of a pharmaceutical factory as an example, this platform manages the operating data of equipment such as fermenters, reactors, and dryers. After receiving equipment data frames, the signal characteristics are analyzed through signal feature recognition algorithms: for the feed valve data of the dryer, it is found that its value range is only 0 and 1, and the state jump is obvious, which is identified as a discrete signal of switching quantity; for the temperature data of the fermenter, it is found that it shows a continuous distribution of 36.5℃-38.2℃ and the change is smooth, which is identified as a continuous change analog signal; for the reaction yield index, it is confirmed that it is calculated from the input and output of raw materials, which is identified as a logic operation calculation signal. For discrete signals such as the feed valve of the dryer, when the state changes from 0 (closed) to 1 (open), the timestamp of the change "2023-06-20T14:30:15.265Z", the state before the change 0, the state after the change 1, and the duration "35 minutes and 20 seconds" (the duration from the last state change to the current state) are recorded and stored in the discrete quantity table of the real-time database. For analog signals like fermenter temperature, a dead zone threshold of 0.2℃ is set. When the temperature changes from 37.1℃ to 37.4℃ (a difference of 0.3℃ exceeding the threshold), a new data point is recorded. However, when the temperature fluctuates between 37.1℃ and 37.2℃ (within the threshold), no data is recorded, thus reducing the original data from one record per second to records that only reflect effective changes. For calculated signals like reaction yield, the input and output parameter IDs, along with the calculation formula "output / input × 100%", are recorded and stored in the calculated data area along with the calculation result (92.5%). As data accumulates, every night at 2 AM, the time-rotation archiver migrates data older than 24 hours to the corresponding date partition in the historical database, while simultaneously calculating the hourly average of each parameter. Finally, the multidimensional index builder creates indexes for all parameters. For example, it creates an index entry with the ID "FT-101" for the fermenter temperature parameter, recording its data type (analog quantity), storage location (real-time table "analog_data" and historical table "hist_analog_2023_06"), and associated temperature control loop "TC-101", forming a complete process parameter index table to support subsequent efficient data access and analysis.

[0056] In one specific embodiment, the process of executing step S103 may specifically include the following steps:

[0057] (1) Retrieve parameter entries with control loop identifiers from the process parameter index table, and extract the control loop dataset with input-output relationship through loop correlation analysis;

[0058] (2) Perform physical connection tracking on the variables in the control loop dataset, use the signal transmission path detector to identify the physical flow path of the signal, and establish a device connection topology diagram;

[0059] (3) Based on the device connection topology, the sequence of signal changes is analyzed by time-series correlation mining technology to identify the trigger chain between control commands and execution responses;

[0060] (4) Based on the triggering chain, the causal inference method is used to quantify the dependence strength between signals and generate a control weight matrix that characterizes the degree of mutual influence between each control link;

[0061] (5) Combine the control weight matrix with the process logic rules, classify the subordinate relationships between control units through a hierarchical organizational structure, and construct a formal control logic tree;

[0062] (6) Systematically integrate and standardize the control logic tree to generate a DCS control model that includes control hierarchy, signal flow, parameter dependency and response time characteristics.

[0063] Specifically, building a DCS control model requires retrieving control loop data from the process parameter index table. This process queries parameter entries with control loop identifiers, such as controller identification codes like "TIC-101" and "FIC-203," which typically follow ISA standard naming rules. Loop correlation analysis extracts control loop datasets with input-output relationships. Loop correlation analysis identifies components belonging to the same control loop based on naming rules and parameter descriptions. For example, for the temperature control loop "TIC-101," associated components include the temperature sensor "TT-101," the controller "TIC-101," and the control valve "TV-101." These associated components form a dataset with clear input-output relationships, reflecting the complete control chain from measurement and control to execution. Physical connection tracing of variables in the control loop dataset is a crucial step in establishing the actual control network. Physical connection tracing determines the physical path of data flow by analyzing the actual connections between devices. Signal transmission path detectors utilize various data sources to identify physical connections between devices, including I / O wiring diagrams, network configuration tables, and signal flow records. The detection process analyzes the data transmission timing, communication protocols, and physical links between different devices to determine the specific transmission path of signals from the source device to the target device. For example, this includes the analog signal transmission path from the sensor to the PLC, and the control signal transmission path from the PLC to the actuator. This physical connection information is integrated to form a device connection topology diagram, which uses nodes to represent devices and edges to represent physical connections, clearly illustrating the physical structure of the industrial control system.

[0064] Based on the device connection topology, time-series correlation mining is used to analyze the causal relationships of signal changes. This technique collects time-series data of signals from various devices, focusing on points of state change. Then, it uses statistical methods to analyze the time interval distribution between different signal change points, identifying regular sequential relationships. For example, it was found that after a change in the controller output signal, the actuator state also changes after a fixed delay, indicating a control relationship between the two. Time-series correlation mining uses a sliding time window method to calculate the conditional probability of signal changes within different time windows, identifying signal pairs with high conditional probabilities. These highly correlated signal pairs form trigger chains, describing the complete trigger path from the initial control command to the final execution response. Based on the trigger chains, causal inference methods quantitatively analyze the dependencies between signals. Causal inference uses statistical methods such as Granger causality tests and transfer entropy calculations to calculate the strength of the dependencies between signals. Granger causality tests assess the contribution of historical data of one signal to predicting the future value of another signal by establishing an autoregressive model; transfer entropy calculation quantifies the degree to which one signal reduces the uncertainty of another signal using information theory methods. These calculations form a control weight matrix, where each element represents the degree of influence between signals. The values ​​are typically between 0 and 1, with larger values ​​indicating stronger dependencies.

[0065] Combining the control weight matrix with process logic rules is the core step in constructing the control logic structure. Process logic rules are equipment operating rules and process control requirements defined based on process knowledge, such as "cooling cannot start simultaneously after heating begins" and "feeding must stop when pressure exceeds limits." The hierarchical organizational structure identifies the master-slave relationships between control units by analyzing the control weight matrix and process rules, classifying interconnected control units according to function and dependency. This process uses a hierarchical clustering algorithm to identify groups of control units with strong dependencies, then determines the hierarchical relationships within the groups according to the control flow, and finally constructs a formalized control logic tree. The control logic tree is a tree-like data structure where the root node typically represents the main control objective, intermediate nodes represent sub-control loops, and leaf nodes represent specific actuators or sensors. The control logic tree is then systematically integrated and standardized to generate a DCS control model. Systematic integration includes merging functionally similar control logic branches, eliminating redundant control paths, and optimizing the control logic structure; standardization converts the control logic into an expression form that conforms to specific standards (such as IEC 61131-3). The final DCS control model contains four key elements: control hierarchy (a multi-layered control structure from the enterprise level to field devices), signal flow (the data transmission path in the control system), parameter dependency (the influence and being influenced relationship between different parameters), and response time characteristics (the time delay characteristics of control signal transmission).

[0066] Taking the distillation tower control system of a petrochemical plant as an example, parameter entries marked "TIC-301" (temperature controller) were retrieved from the process parameter index table. Loop correlation analysis confirmed that the control loop included a temperature sensor TT-301, a control calculation unit TIC-301, and a steam regulating valve TV-301. Physical connection tracing revealed that the temperature signal was transmitted from the sensor to the DCS I / O card via a 4-20mA current signal, then to the controller via the internal bus. The control output was transmitted to the regulating valve via another I / O card as a 4-20mA signal, forming a closed-loop control loop. Time-series correlation mining analysis of historical data revealed that when the temperature sensor TT-301 showed a 1°C temperature increase, approximately 2 seconds later, the controller TIC-301 output decreased by 5%. One second later, the regulating valve TV-301 opened less, and approximately 15 seconds later, the temperature began to decrease, forming a complete trigger chain. Causal inference determined that the weight of the controller output on the valve position was 0.92, and the weight of the temperature on the controller output was 0.85, forming a control weight matrix. By combining process rules such as "temperature control has higher priority than flow control", a control logic tree is constructed, determining that the temperature control of the distillation column is the main loop and the flow control is the subordinate loop. Finally, a DCS control model is formed, which clearly describes the complete control process from temperature measurement and PID calculation to valve adjustment, including key characteristics such as a 2-second calculation delay and a 15-second process response time.

[0067] In one specific embodiment, the process of executing step S104 may specifically include the following steps:

[0068] (1) Extract key process node parameters from the DCS control model, and construct parameter change trend feature vectors using the time series piecewise linear regression method to form a dynamic feature library of process parameters;

[0069] (2) The dynamic feature library of process parameters is associated and mapped with the equipment operating status data, and the abnormal fluctuation pattern of parameters is identified by hierarchical clustering algorithm to generate a process fluctuation feature map;

[0070] (3) Apply the principal component anomaly measurement method to the process fluctuation characteristic spectrum for multidimensional spatial anomaly detection, and determine the degree of parameter deviation from normal operating conditions by Mahalanobis distance calculation to form a quantitative index of process deviation;

[0071] (4) Based on the quantitative indicators of process deviation, an alarm association network is constructed, and the root cause alarm and propagation path are determined by Bayesian causal inference analysis to form an alarm propagation tree;

[0072] (5) Extract the root cause fault point from the alarm propagation tree, perform similarity matching with the equipment historical fault case library, calculate the probability distribution of fault type through fuzzy comprehensive evaluation method, and generate the abnormal cause classification result;

[0073] (6) Combine the results of the abnormal cause classification with the process parameter fluctuation characteristics for analysis, and generate a process abnormality diagnosis report containing abnormal description, root cause analysis, impact degree and treatment suggestions through the decision rule engine.

[0074] Specifically, key process node parameters are extracted from the DCS control model. Key process node parameters refer to core parameters that significantly impact process control and quality, including control loop setpoints, measured values, output values, and key process variables. After extracting these parameters, a parameter change trend feature vector is constructed using time-series piecewise linear regression. Time-series piecewise linear regression is a technique that divides continuous time-series data into several linear segments. Inflection points are determined using a sliding window or adaptive segmentation algorithm, and then a linear model is fitted to each segment, extracting features such as slope, intercept, and fitting error. These features collectively form a trend feature vector, describing the rate of change, stability, and fluctuation characteristics of parameters over different time periods, and are stored in a dynamic feature library of process parameters. Associating and mapping the dynamic feature library of process parameters with equipment operating status data is fundamental to identifying abnormal patterns. Equipment operating status data includes equipment start / stop status, operating load, vibration monitoring data, and electrical parameters. These equipment data are aligned with parameter features using timestamps to establish a parameter-equipment status correlation matrix. Hierarchical clustering algorithms are methods for constructing hierarchical data structures from bottom to top or top to bottom. In process parameter clustering, Euclidean distance or dynamic time warping distance is typically chosen as a similarity measure, and Ward's minimum variance method is used as the clustering merging criterion. The clustering process groups similar parameter fluctuation patterns together, forming fluctuation feature groups that characterize different anomaly types, ultimately generating a process fluctuation feature map that reflects the global process status.

[0075] Principal component anomaly measurement involves performing principal component analysis on multidimensional parameter data, transforming the raw data into a principal component space, and then calculating the Mahalanobis distance from the sample points to the center of normal operating conditions in the principal component space to determine the degree of anomaly. The specific calculation formula is as follows:

[0076]

[0077] in, This represents the Mahalanobis distance of sample point x. It is a feature vector containing multiple process parameters. It is the average eigenvector under normal operating conditions. It is the covariance matrix It is the inverse of the covariance matrix. For data after principal component analysis, the Mahalanobis distance calculation can be simplified to:

[0078]

[0079] here, It is based on the Mahalanobis distance of the principal components. It is the score of the sample on the i-th principal component. It is the eigenvalue of the i-th principal component. It is the number of principal components retained. It is the original feature dimension. It is a reconstruction error. These are adjustment parameters. The first term reflects the degree of deviation of the sample in the primary direction of variation, and the second term reflects the degree of deviation of the sample in the secondary direction of variation. These are adjusted by setting a threshold. ,when When an abnormality is detected, a quantitative indicator of process deviation is formed.

[0080] An alarm association network is a graph structure representing the causal relationships between alarm events. Nodes represent alarm events for each parameter, and edges represent the propagation relationships between alarms. Bayesian causal inference analysis quantitatively assesses the causal strength between alarms by calculating conditional probabilities, establishing a Bayesian network structure between parameters, learning the conditional probability distribution using historical data, and finally determining the most probable causal chain under given observation conditions through posterior probability inference. This process forms an alarm propagation tree, clearly showing the complete propagation path from the root alarm to the chain reaction. Extracting the root fault point from the alarm propagation tree is a fundamental step in determining the nature of the fault. The root fault point is the root node of the alarm propagation tree, i.e., the original fault that triggers other chain alarms. After extracting these root fault points, they are matched with a historical fault case database. The historical fault case database contains records of past faults, each record including parameter characteristics at the time of the fault, alarm sequence, fault type, and handling method. Similarity matching adopts fuzzy comprehensive evaluation method. This method establishes multiple fuzzy relation matrices to represent the membership degree of fault cases under different evaluation indicators. Then, the comprehensive evaluation result is obtained through fuzzy synthesis operation, the matching probability of the current fault and the historical fault type is calculated, and the abnormal cause classification result is generated.

[0081] The final step in generating a diagnostic report is to combine the anomaly cause classification results with the process parameter fluctuation characteristics for analysis. The decision rule engine constructs a series of IF-THEN rules based on expert knowledge and historical experience to assess the nature, severity, and scope of impact of the anomaly. The rule structure typically includes a condition part (IF) and a conclusion part (THEN). The condition part checks the anomaly type and parameter deviation, while the conclusion part provides the diagnostic results and handling recommendations. Through the rule reasoning engine, a process anomaly diagnostic report is generated, containing anomaly description (anomaly parameters, deviation values, occurrence time), root cause analysis (fault source, fault mechanism, propagation path), impact level (quality impact, safety risk, production loss), and handling recommendations (adjustment plan, maintenance items, emergency measures). Taking a petrochemical plant as an example, the plant's heater experienced temperature fluctuations. Key process parameters such as heater outlet temperature, fuel gas flow rate, and oxygen content were extracted from the DCS control model. Time series piecewise linear regression analysis revealed a continuous upward trend in outlet temperature, with the slope increasing from the normal 0.5℃ / hour to 2.3℃ / hour, accompanied by periodic fluctuations of approximately 15 minutes and an amplitude of ±4℃. After associating these features with equipment operating status data, a hierarchical clustering algorithm identified a specific fluctuation pattern: temperature rise accompanied by increased fuel gas flow and decreased oxygen content. This pattern is highly similar to "combustion control anomalies" in historical data. Principal component anomaly measurement was applied to analyze this fluctuation feature map. Mahalanobis distance calculation revealed a normalized distance value of 4.8 for the current operating condition, significantly higher than the threshold of 2.5 for normal operating conditions, indicating a significant deviation. After constructing an alarm association network, Bayesian causal inference analysis determined the alarm sequence: combustion air flow sensor drift caused abnormal oxygen content control, which in turn triggered fuel gas flow fluctuations, ultimately leading to unstable outlet temperature. The root cause fault point extracted from the alarm propagation tree was "FT-103 combustion air flow sensor drift." Matching this with the historical fault case database revealed the highest similarity (87%) to the case of "pressure tap blockage causing low sensor readings." The final process anomaly diagnosis report indicated that the root cause of the abnormal fluctuation in the furnace outlet temperature was the low reading of the combustion air flow sensor due to a blockage at the pressure tap, resulting in insufficient response of the control system, insufficient actual air volume, and decreased combustion efficiency. It was recommended to clean the sensor pressure tap and temporarily increase the air-fuel ratio setting to 1.15, and to arrange for the sensor assembly to be replaced during the next shutdown.

[0082] In one specific embodiment, the process of executing step S105 may specifically include the following steps:

[0083] (1) Extract control loop anomaly information from process anomaly diagnosis report, accurately locate the affected PID control loops through fault impact analysis matrix, and form a list of control loops to be adjusted;

[0084] (2) For each control loop in the list of control loops to be adjusted, extract the process response curves of normal and abnormal operating conditions based on the historical database, construct the process dynamic model through system identification technology, and generate the transfer function of the control object.

[0085] (3) Input the control object transfer function into the improved Ziegler-Nichols algorithm, calculate the initial PID parameter set through the critical oscillation method, and form the parameter initial value matrix;

[0086] (4) Perform multi-objective optimization on the initial parameter matrix. Using the particle swarm optimization algorithm combined with the ITAE performance index minimization principle, iteratively calculate the optimal combination of the proportional, integral, and derivative parameters to generate the optimal PID parameter scheme.

[0087] (5) Based on the optimal parameter scheme of PID, the robustness analysis of the control system is carried out by Monte Carlo simulation technology to generate a safe operating envelope that includes the stable operating range and the boundary of the extreme operating conditions.

[0088] (6) Compare and verify the safe operation envelope with the process safety specifications, and confirm that the control parameters meet the safety boundary requirements through the safety margin evaluator to form the actuator control command that has been verified by safety.

[0089] Specifically, the PID parameter self-tuning process of the automated control platform's data management method extracts control loop anomaly information from the process anomaly diagnosis report. Text mining techniques are used to analyze the report content, identifying keywords and phrases related to control anomalies, such as "slow controller response," "lagging control valve action," and "periodic increase in temperature fluctuations." Then, a fault impact analysis matrix is ​​used to precisely locate the affected PID control loops. This matrix is ​​a two-dimensional table where the horizontal axis represents various fault symptoms, and the vertical axis represents each control loop. Matrix element values ​​reflect the degree of impact of the fault symptoms on the control loops. During the analysis, the symptoms identified in the diagnosis report are mapped to the horizontal axis of the matrix, and control loops with impact values ​​exceeding a preset threshold are extracted, forming a list of control loops to be adjusted. This list includes the control loop ID, fault impact score, and preliminary diagnostic results. For each control loop in the list, a process response curve needs to be extracted from a historical database. The process response curve refers to the dynamic response trajectory of the control variable when the input changes, including both normal and abnormal operating condition response curves. The extraction method involves retrieving setpoint change records related to the target control loop from a historical database, extracting process variables and control output data over a period before and after the change, and forming a standard step response dataset. A process dynamic model is constructed using system identification technology. System identification is a method for deriving a mathematical model of a system from input-output data, including two main categories: parametric model identification and non-parametric model identification. In this method, the least squares method is used to fit a first-order model with dead zone or a second-order underdamped model to obtain key parameters such as process gain, time constant, and pure delay, generating the control object transfer function. The control object transfer function is a mathematical model describing the input-output relationship of a control system, typically expressed as a fractional form in the Laplace domain.

[0090] The control object transfer function is input into the improved Ziegler-Nichols algorithm, and the initial PID parameter set is calculated using the critical oscillation method. The improved Ziegler-Nichols algorithm is an optimized version of the classic Ziegler-Nichols method, and the critical oscillation method is its core step. The specific calculation formula is as follows:

[0091]

[0092] in, It is proportional gain. It is the integration time constant. It is the differential time constant. It is the critical gain (the minimum gain value that makes the system produce constant amplitude oscillations). It is the critical oscillation period. It is the system's pure delay time. , , , , and These are adjustment coefficients for the improved algorithm, used to optimize the parameter calculations of the classic Ziegler-Nichols method. These coefficients are typically preset based on the type of controlled object; for example, for a slow temperature control system, =0.33、 =0.5、 =1.2、 =0.6、 =0.8、 =0.3; for fast flow control systems, these coefficients will have different values. The PID parameters calculated using this set of formulas form an initial parameter matrix, containing multiple parameter combinations under different adjustment coefficients. Multi-objective optimization is performed on the initial parameter matrix, using a particle swarm optimization algorithm combined with the ITAE (Integral Time Absolute Error) performance index minimization principle to iteratively calculate the optimal parameter combination. The particle swarm optimization algorithm is a swarm intelligence optimization method that simulates the foraging behavior of bird flocks. Each particle represents a set of PID parameter combinations, moving in the parameter space to find the optimal solution. During algorithm initialization, each particle is randomly assigned a position (parameter value) and velocity, and then the position and velocity of each particle are iteratively updated, moving towards the individual optimal position and the group optimal position. The ITAE performance index is a mathematical metric for evaluating the quality of a control system, calculating the integral of the absolute value of the control error multiplied by time, with particular attention to errors existing over long periods. The optimization process sets minimizing the ITAE as the objective function, while incorporating multiple constraints such as overshoot, rise time, and settling time. Through iterative calculation, the optimal PID parameter scheme balancing various performance indices is finally obtained. Based on the optimal PID parameter scheme, the robustness analysis of the control system is performed using Monte Carlo simulation technology. Monte Carlo simulation is a statistical method for evaluating system behavior through extensive random sampling. In the robustness analysis of control systems, it identifies key uncertain parameters and their distribution characteristics, such as the process gain fluctuation range, time constant variation range, and external disturbance amplitude. Then, a large number of random simulations are performed, randomly assigning values ​​to these uncertain parameters in each simulation. Closed-loop control system simulations are executed, and key indicators are recorded. By statistically analyzing these simulation results, the performance boundaries of the system under parameter variations are determined, generating a safe operating envelope that includes the stable operating range and extreme operating condition boundaries. The safe operating envelope is a two-dimensional or multi-dimensional graph representing the boundary conditions for stable operation of the control system under different parameter combinations. Process safety specifications are operating limits established based on equipment physical limitations, process technology requirements, and safety management regulations, typically including upper and lower pressure limits, temperature safety boundaries, and allowable flow ranges. A safety margin evaluator is a tool for analyzing the safety redundancy of a control system. It calculates the distance between key parameters (such as maximum pressure and maximum temperature) and safety limits under worst-case conditions, forming a safety margin index. The evaluator checks indicators such as the stability boundaries, overshoot limits, and fluctuation frequencies of control parameters to ensure that safety thresholds are not triggered under various conditions. Only control parameters that simultaneously meet stability and safety boundary requirements will be accepted and form safety-verified actuator control instructions, including PID parameter setpoints, control mode selection, and operating range limits.

[0093] In one specific embodiment, the process of executing step S106 may specifically include the following steps:

[0094] (1) Collect real-time data of process parameters after the actuator control command is implemented, and measure the settling time and overshoot before and after the system adjustment through the response time analyzer to form a set of control effect evaluation indicators;

[0095] (2) Compare the control effect evaluation index set with the preset production target index, calculate the degree of deviation of each index by weighted Euclidean distance, and generate a multidimensional deviation vector;

[0096] (3) Based on the multidimensional deviation vector, sensitivity analysis is performed on the key formulation parameters in the process formulation library. The key formulation factors that have a significant impact on product quality and production efficiency are identified by orthogonal experimental design method, and a formulation parameter sensitivity matrix is ​​constructed.

[0097] (4) For the high-sensitivity factors in the formula parameter sensitivity matrix, the adaptive incremental learning algorithm is used to iteratively optimize the process formula. By extracting the parameter combination features of historical successful cases, a process formula optimization and adjustment scheme is formed.

[0098] (5) Link and store the process formula optimization and adjustment scheme with the actual production effect data, and use knowledge graph construction technology to express the mapping relationship between process parameters, equipment status, environmental conditions and product quality in a structured way to form a production knowledge network;

[0099] (6) Based on the production knowledge network, a recommended solution based on historical successful experience is provided for the setting of process parameters under specific production conditions through similar scenario retrieval engine and pattern matching technology, and an intelligent production recommendation knowledge base is constructed.

[0100] Specifically, the final stage of the automated control platform's data management method involves process optimization and knowledge base construction, requiring the collection of real-time process parameter data after the actuator control commands are implemented. This data is acquired through a high-frequency acquisition mechanism of a real-time database, including process variables such as temperature, pressure, and flow rate, as well as product quality indicators. A response time analyzer is a data processing tool specifically designed for evaluating control system performance. It calculates key performance indicators by analyzing the time-series data of control variables. Specifically, the response time analyzer identifies the time points when system parameters change, then tracks the parameter change trajectory, extracting key time nodes and numerical characteristics. Settling time refers to the time required for the system to stabilize near the target value after a disturbance, typically defined as the moment when the parameter enters and remains within ±5% of the target value; overshoot refers to the maximum deviation of the parameter value from the target value, usually expressed as a percentage. These indicators, along with rise time, settime, and steady-state error, constitute the control effectiveness evaluation indicator set, comprehensively reflecting the dynamic performance and steady-state accuracy of the control system. Comparing the control effectiveness evaluation indicator set with preset production target indicators is the core step in evaluating control effectiveness. The preset production target indicators are ideal control performance parameters set based on product quality requirements and production efficiency goals, including maximum permissible overshoot, expected settling time, and upper and lower limits of product quality indicators. Weighted Euclidean distance is a method for measuring the distance between points in a multidimensional space; in this scheme, it is used to quantify the deviation between the actual control effect and the target requirements. During the calculation, different indicators are assigned different weights according to their importance; for example, product quality indicators usually have higher weights, while process indicators such as settling time have relatively lower weights. The calculation results form a multidimensional deviation vector, where each component represents the degree of deviation of a specific indicator, and the vector length reflects the distance between the overall control effect and the target.

[0101] Sensitivity analysis of key formulation parameters in a process formulation library based on multidimensional deviation vectors is fundamental to optimizing process formulations. The process formulation library is a database storing different product formulations and process parameters, including raw material ratios, reaction conditions, and processing parameters. Sensitivity analysis employs orthogonal experimental design, a highly efficient multi-factor experimental design technique that can analyze the influence of multiple factors with a relatively small number of experiments. Orthogonal experimental design determines the formulation parameters to be analyzed and their respective value levels. Then, different parameter combinations are arranged according to an orthogonal array, experiments are conducted, and results are recorded. Range analysis and variance analysis are used to calculate the degree of influence of different parameters on the results, identifying key formulation factors that significantly affect product quality and production efficiency. These key factors and their degree of influence constitute a formulation parameter sensitivity matrix, providing guidance for subsequent formulation optimization. For highly sensitive factors in the formulation parameter sensitivity matrix, iterative optimization using an adaptive incremental learning algorithm is required. Adaptive incremental learning is a machine learning method that continuously absorbs new knowledge and adjusts its learning strategy, making it particularly suitable for parameter optimization scenarios in industrial processes. The algorithm extracts parameter combination features from historical data of successful cases, identifying formulation parameter patterns corresponding to high-quality products. Then, based on the current production conditions and the multidimensional deviation vector, the direction and step size of parameter adjustment are calculated, generating new parameter combination suggestions. This process is iterative; after each parameter adjustment, the system collects new production effect data, updates the parameter-effect mapping model, and further optimizes the parameter combination. Through multiple iterations, a process formulation optimization and adjustment scheme tailored to specific production conditions is formed.

[0102] Linking and storing process formulation optimization and adjustment plans with actual production performance data is a crucial step in building a knowledge base. This linked storage not only records the final optimized formulation parameters but also includes intermediate results and corresponding production effects during the adjustment process, forming a complete optimization history. Knowledge graph construction technology transforms this linked data into a structured knowledge network. A knowledge graph is a semantic network structure representing entities and their relationships. In the field of industrial control, entities include process parameters, equipment status, environmental conditions, and product quality, while relationships describe the dependencies, influences, and constraints between them. The construction process includes three main steps: entity identification, relationship extraction, and knowledge fusion, ultimately forming a semantic network reflecting industrial production knowledge, known as the production knowledge network.

[0103] Building an intelligent production recommendation knowledge base based on a production knowledge network is the final step in achieving knowledge reuse. A similar scenario retrieval engine is a search tool capable of quickly finding scenarios similar to the current situation from a large number of historical cases. Its core is a similarity calculation algorithm, typically using methods such as cosine similarity and Jaccard coefficient to calculate the degree of similarity between scenarios. Pattern matching technology is responsible for extracting reusable parameter settings and operating methods from similar scenarios. By analyzing the common characteristics of successful cases, it identifies best practices applicable to the current production conditions. The intelligent production recommendation knowledge base not only stores these recommended solutions but also records the applicable conditions, historical success rate, and source of each solution, facilitating users' evaluation of the reliability of the recommendations. Taking a batch reaction process in a fine chemical plant as an example, the process optimization and knowledge base construction process are illustrated. This plant produces a high-purity catalyst, involving precise temperature control and formula adjustment. After optimizing the PID parameters and issuing actuator control commands, high-precision temperature sensors collect reactor temperature data, revealing that the temperature control overshoot decreased from 3.5℃ to 1.8℃, and the stabilization time shortened from 42 minutes to 25 minutes. The response time analyzer evaluates the control effect and calculates indicators such as temperature stabilization time, overshoot, and control accuracy to form a set of control effect evaluation indicators. These indicators are compared with preset targets (overshoot < 2.0℃, stabilization time < 30 minutes, control accuracy ± 0.5℃). Through weighted Euclidean distance calculation, a multidimensional deviation vector [0.2, -5.0, 0.1] is obtained, indicating that the overshoot is slightly exceeded, the stabilization time is better than the target, and the control accuracy basically meets the requirements. Based on this deviation vector, sensitivity analysis is performed on parameters such as reaction temperature, catalyst addition rate, and stirring speed in the process formulation. Nine experimental schemes are designed using an L9(34) orthogonal array. Through range analysis, it is determined that the catalyst addition rate and reaction temperature are the two most significant factors affecting product purity, with sensitivities of 0.85 and 0.68, respectively. For these two highly sensitive factors, an adaptive incremental learning algorithm analyzed the parameter-purity relationship in historical production data from 60 batches, suggesting a slight adjustment to the catalyst addition rate from 25 kg / h to 23.5 kg / h and the reaction temperature from 165℃ to 168℃. This optimization suggestion was linked and stored with actual production performance data, showing a 0.8 percentage point increase in product purity and a 7% reduction in energy consumption. Through knowledge graph technology, the four entities—catalyst addition rate, reaction temperature, product purity, and energy consumption—and their relationships were structured, incorporating influencing factors such as environmental temperature and humidity, and raw material batches, forming a complete production knowledge network.Subsequently, when the factory handles similar batch production, it quickly finds matching historical success cases through a similar scenario search engine and intelligently recommends process parameter settings suitable for the current conditions. For example, it suggests reducing the reaction temperature by 2°C in high-temperature environments in summer, or increasing the stirring speed by 15% when handling high-viscosity raw materials. These recommendations are stored in the intelligent production recommendation knowledge base, continuously providing decision support for subsequent production.

[0104] The above describes the method for managing the operation data of the automation control platform in the embodiments of this application. The following describes the management system for the operation data of the automation control platform in the embodiments of this application. Please refer to [link / reference]. Figure 2 One embodiment of the automated control platform operation data management system in this application includes:

[0105] The acquisition module is used to acquire the operating status and process parameters of industrial equipment with the industrial controller through the fieldbus network, and generate industrial data frames containing control point location information after edge processing by the PLC.

[0106] The storage module is used to classify and store the industrial data frames into the real-time database and the historical database according to the signal type of discrete quantities, analog quantities and computational quantities, and to establish a process parameter index table.

[0107] The module is used to extract control loop data from the process parameter index table, construct the signal flow direction and control dependency relationship between equipment, and form a DCS control model;

[0108] The generation module is used to generate a process anomaly diagnosis report based on the DCS control model, combined with the process parameter change trend and equipment operating status, through multi-dimensional alarm analysis technology.

[0109] The calculation module is used to calculate the optimal control parameters based on the process anomaly diagnosis report using a PID parameter self-tuning algorithm, while checking the operation safety boundary and outputting actuator control commands.

[0110] The analysis module is used to analyze the deviation between the implementation effect of the actuator control commands and the production target, dynamically optimize the process formula library, and build an intelligent production recommendation knowledge base.

[0111] Through the collaborative efforts of the aforementioned components, the system collects the operating status and process parameters of industrial equipment via a fieldbus network and industrial controller. After edge processing by the PLC, it generates industrial data frames containing control point information, achieving unified data acquisition and standardized processing of distributed industrial equipment, thus improving the completeness and accuracy of data acquisition. Based on the signal type of the industrial data frames, they are categorized into discrete, analog, and computational quantities and stored in real-time and historical databases, respectively, and a process parameter index table is established. A differentiated data storage strategy is adopted, significantly optimizing storage efficiency and improving data retrieval performance. Control loop data is extracted from the process parameter index table to construct the signal flow and control dependencies between devices, forming a DCS control model. This enables the system to comprehensively understand the topology and data flow paths of the industrial control system, providing a system-level perspective for subsequent analysis. Based on the DCS control model, combined with process parameter change trends and equipment operating status, a process anomaly diagnostic report is generated through multi-dimensional alarm analysis technology. This process applies artificial intelligence algorithms for multi-dimensional data fusion and correlation analysis, ensuring that the diagnostic results are not limited to single-point threshold judgments but analyzed from the perspective of overall system behavior, significantly improving efficiency. The accuracy of anomaly detection and the precision of root cause analysis, especially the application of algorithms such as Bayesian causal inference and principal component anomaly measurement, enable the system to accurately identify the root causes of anomalies and trace their propagation paths in complex industrial environments. Based on the process anomaly diagnosis report, the system uses a PID parameter self-tuning algorithm to calculate the optimal control parameters, while simultaneously verifying the operational safety boundary and outputting actuator control commands. In this step, the combined application of the improved Ziegler-Nichols algorithm and particle swarm optimization algorithm fully leverages the respective advantages of traditional control theory and modern optimization algorithms, achieving automatic optimization of control parameters and ensuring safety, significantly reducing the workload of manual parameter tuning and improving control accuracy. The system then performs deviation analysis between the implementation effect of actuator control commands and production targets, dynamically optimizes the process formula library, and constructs an intelligent production recommendation knowledge base. This step, through adaptive incremental learning algorithms and knowledge graph technology, establishes a complete knowledge accumulation and experience inheritance mechanism. The system can continuously learn from production practice and optimize process parameters, forming a virtuous cycle of continuous improvement. This achieves the structured expression and efficient reuse of industrial knowledge, providing data-driven intelligent recommendation support for decision-making under similar production conditions.

[0112] Reference Figure 3 This invention also provides a computer device, which can be a server, and its internal structure can be as follows: Figure 3As shown, the computer device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores the data corresponding to this embodiment. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0113] Those skilled in the art will understand that Figure 3 The structures shown are merely block diagrams of some structures related to the present invention and do not constitute a limitation on the computer devices on which the present invention is applied.

[0114] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0115] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the present invention and embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0116] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0117] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0118] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for managing operational data of an automated control platform, characterized in that, The automated control platform operation data management method includes: The PLC collects the operating status and process parameters of industrial equipment with the industrial controller via a fieldbus network, and generates industrial data frames containing control point information after edge processing. Based on the signal type of the industrial data frame, it is classified into discrete quantities, analog quantities, and computational quantities and stored in the real-time database and historical database to establish a process parameter index table. Extract control loop data from the process parameter index table, construct the signal flow direction and control dependency relationship between equipment, and form a DCS control model; Based on the DCS control model, and combining process parameter change trends and equipment operating status, a process anomaly diagnosis report is generated using multi-dimensional alarm analysis technology. This includes: extracting key process node parameters from the DCS control model; constructing parameter change trend feature vectors using time-series piecewise linear regression to form a dynamic feature library of process parameters; mapping the dynamic feature library of process parameters to equipment operating status data; identifying abnormal parameter fluctuation patterns using a hierarchical clustering algorithm to generate a process fluctuation feature map; and applying principal component anomaly measurement to the process fluctuation feature map for multi-dimensional spatial anomaly detection, using a Mahalanobis distance meter. The degree to which parameters deviate from normal operating conditions is determined to form a quantitative index of process deviation. Based on the quantitative index of process deviation, an alarm association network is constructed, and the root cause alarm and propagation path are determined through Bayesian causal inference analysis to form an alarm propagation tree. The root cause fault point is extracted from the alarm propagation tree, and similarity matching is performed in combination with the equipment's historical fault case library. The probability distribution of fault type is calculated through fuzzy comprehensive evaluation to generate anomaly cause classification results. The anomaly cause classification results are combined with process parameter fluctuation characteristics for analysis, and a process anomaly diagnosis report containing anomaly description, root cause analysis, impact degree, and handling suggestions is generated through a decision rule engine. Based on the process anomaly diagnosis report, the optimal control parameters are calculated using a PID parameter self-tuning algorithm, while simultaneously verifying the operational safety boundary, and outputting actuator control commands. This includes: extracting control loop anomaly information from the process anomaly diagnosis report; accurately locating the affected PID control loops using a fault impact analysis matrix to form a list of control loops to be adjusted; for each control loop in the list, extracting process response curves for normal and abnormal operating conditions from a historical database; constructing a process dynamic model using system identification technology to generate a control object transfer function; and inputting the control object transfer function into an improved Ziegler-Nichols algorithm, which, through fault... The initial PID parameter set is calculated using the boundary oscillation method to form an initial parameter value matrix. Multi-objective optimization is then performed on this initial parameter value matrix. The optimal combination of proportional, integral, and derivative parameters is iteratively calculated using a particle swarm optimization algorithm combined with the ITAE performance index minimization principle to generate the optimal PID parameter scheme. Based on this optimal PID parameter scheme, robustness analysis of the control system is performed using Monte Carlo simulation technology to generate a safe operating envelope containing the stable operating range and extreme condition boundaries. This safe operating envelope is then compared and verified with process safety specifications. A safety margin evaluator confirms that the control parameters meet the safety boundary requirements, resulting in a safety-verified actuator control command. The deviation analysis between the implementation effect of the actuator control commands and the production target is performed, the process formula library is dynamically optimized, and an intelligent production recommendation knowledge base is constructed.

2. The automated control platform operation data management method according to claim 1, characterized in that, The process involves acquiring the operating status and process parameters of industrial equipment via a fieldbus network and an industrial controller, and then generating an industrial data frame containing control point information after edge processing by the PLC. This includes: The distributed fieldbus interface is used to connect the communication ports of various industrial devices to collect raw process parameters, including temperature, pressure, flow rate, position, and switch status. Key control point data are selected from the original process parameters, and high-frequency changing parameters are sampled at the millisecond level and low-frequency changing parameters are sampled at the minute level through a sampling frequency adaptive adjustment mechanism. The sampled data is introduced into the PLC edge computing unit, and the data in different formats is unified into the target standard format through a data standardization conversion protocol. The target standard format data is filtered for signal and noise, and abnormal fluctuation values ​​are removed using a sliding window midpoint filtering algorithm. Add timestamps and device identifiers to the filtered data, and construct the metadata header area and data payload area; The metadata header area and the data payload area are packaged and integrated into an industrial data frame containing control point location information, and the data is securely protected through a lightweight encryption protocol.

3. The automated control platform operation data management method according to claim 1, characterized in that, The process parameter index table is established by classifying the industrial data frames according to signal type (discrete, analog, and computational) and storing them in a real-time and historical database, and by creating a process parameter index table. The industrial data frame is parsed for signal type, and the input signal is divided into discrete signals of switching quantity, analog signals of continuous change, and calculation signals of logic operation through signal feature recognition algorithm; For the aforementioned discrete signals of the switching quantity type, a state change triggering recording mechanism is adopted to store the state transition point data into the discrete quantity data area of ​​the real-time database, while retaining the state duration information; For the continuously changing analog signals, the data compression engine performs dead-zone compression on the original signals, and writes the data points into the analog data area of ​​the real-time database only when the signal changes exceed a set threshold. For the aforementioned logical operation-type computation signals, the input source and computation formula are recorded by a computation dependency tracker, and the computation result and formula are stored together in the computational data area of ​​the real-time database. Based on the data lifecycle strategy, the time-rotation archiver periodically transfers historical data from the real-time database to the historical database and organizes the data structure according to the time-partitioned storage method; The multidimensional index builder establishes an access path mapping for all process parameters in the real-time database and the historical database, generating a process parameter index table containing parameter ID, data type, storage location, and associated control loop information.

4. The automated control platform operation data management method according to claim 1, characterized in that, The step of extracting control loop data from the process parameter index table, constructing the signal flow and control dependency relationship between equipment, and forming a DCS control model includes: Retrieve parameter entries with control loop identifiers from the process parameter index table, and extract control loop datasets with input-output relationships through loop correlation analysis; Physical connection tracking is performed on the variables in the control loop dataset, and the physical flow path of the signal is identified using a signal transmission path detector to establish a device connection topology. Based on the device connection topology, the sequence of signal changes is analyzed using time-series correlation mining technology to identify the trigger chain between control commands and execution responses. Based on the triggering chain, the causal inference method is used to quantify the dependence strength between signals and generate a control weight matrix that characterizes the degree of mutual influence between each control link. By combining the control weight matrix with process logic rules, a formalized control logic tree is constructed by classifying the subordinate relationships between control units through a hierarchical organizational structure. The control logic tree is systematically integrated and standardized to generate a DCS control model that includes control hierarchy, signal flow, parameter dependencies, and response time characteristics.

5. The automated control platform operation data management method according to claim 1, characterized in that, The process of analyzing the deviation between the implementation effect of the actuator control commands and the production target, dynamically optimizing the process formula library, and constructing an intelligent production recommendation knowledge base includes: Real-time data of process parameters after the actuator control command is implemented are collected, and the settling time and overshoot before and after system adjustment are measured by the response time analyzer to form a set of control effect evaluation indicators. The control effect evaluation index set is compared with the preset production target index, and the deviation degree of each index is calculated by weighted Euclidean distance to generate a multi-dimensional deviation vector. Based on the multidimensional deviation vector, sensitivity analysis is performed on key formulation parameters in the process formulation library. Key formulation factors that have a significant impact on product quality and production efficiency are identified through orthogonal experimental design, and a formulation parameter sensitivity matrix is ​​constructed. For the high-sensitivity factors in the formula parameter sensitivity matrix, an adaptive incremental learning algorithm is used to iteratively optimize the process formula. By extracting the parameter combination features of historical successful cases, a process formula optimization and adjustment scheme is formed. The process formula optimization and adjustment scheme is associated with and stored with actual production effect data. The mapping relationship between process parameters, equipment status, environmental conditions and product quality is expressed in a structured way through knowledge graph construction technology, forming a production knowledge network. Based on the aforementioned production knowledge network, a similar scenario retrieval engine and pattern matching technology are used to provide recommended solutions for setting process parameters under specific production conditions based on historical successful experiences, thereby constructing an intelligent production recommendation knowledge base.

6. An automated control platform operation data management system, used to implement the automated control platform operation data management method as described in any one of claims 1 to 5, characterized in that, The automated control platform's operational data management system includes: The acquisition module is used to acquire the operating status and process parameters of industrial equipment with the industrial controller through the fieldbus network, and generate industrial data frames containing control point location information after edge processing by the PLC. The storage module is used to classify and store the industrial data frames into the real-time database and the historical database according to the signal type of discrete quantities, analog quantities and computational quantities, and to establish a process parameter index table. The module is used to extract control loop data from the process parameter index table, construct the signal flow direction and control dependency relationship between equipment, and form a DCS control model; The generation module is used to generate a process anomaly diagnosis report based on the DCS control model, combined with the process parameter change trend and equipment operating status, through multi-dimensional alarm analysis technology. The calculation module is used to calculate the optimal control parameters based on the process anomaly diagnosis report using a PID parameter self-tuning algorithm, while checking the operation safety boundary and outputting actuator control commands. The analysis module is used to analyze the deviation between the implementation effect of the actuator control commands and the production target, dynamically optimize the process formula library, and build an intelligent production recommendation knowledge base.

7. A computer device, characterized in that, The system includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the automated control platform operation data management method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, the computer program causing a processor, when executed by a processor, to perform the automated control platform operation data management method as described in any one of claims 1 to 5.

Images