Data processing method and system of chemical production information management platform

Abstract

The present application relates to the technical field of data processing, in particular to a data processing method and system of a chemical production informatization management platform, in the present application, real-time data of chemical production is acquired, and a production context label representing a current production situation is generated; according to the production context label, a corresponding parameter configuration set is dynamically selected, and the parameter configuration set is used for setting operation parameters of an abnormality detection algorithm; according to the production context label, a data subset associated with the production context label in time is selected from the real-time data of chemical production, and the selected data subset is subjected to correlation analysis; the result of the correlation analysis is operated using the abnormality detection algorithm set by the parameter configuration set, and a state indication signal is output; through context perception, dynamic optimization of the detection strategy is realized, and the problem of poor working condition adaptability caused by fixed parameters and global analysis in the traditional method is solved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a data processing method and system for an information management platform for chemical production. Background Technology

[0002] The information management platform for chemical production is a core system for achieving intelligent operation in the modern chemical industry. It collects, processes, and analyzes real-time data generated during production to monitor production status, provide early warnings of anomalies, and support decision-making. In chemical production, the process typically involves complex technological chains, variable operating conditions, and highly integrated equipment systems. Therefore, real-time data is characterized by massive volume, heterogeneity, and strong temporal sequence. Traditional anomaly detection methods usually rely on preset fixed parameter sets and static data analysis models. These models are often trained based on historical data or ideal operating conditions, failing to fully consider the dynamic changes in the production context. For example, during production, different production stages (such as startup, steady-state operation, and shutdown) or equipment states (such as normal, maintenance, and fault) can cause significant changes in data distribution and the relationships between variables. Fixed-parameter anomaly detection algorithms struggle to adapt to these changes, easily leading to false alarms or missed alarms.

[0003] Furthermore, traditional methods for data correlation analysis often employ global or fixed time windows, neglecting the temporal characteristics of the production context. This leads to a disconnect between the analysis results and the actual production status, reducing the accuracy and timeliness of anomaly detection. This limitation prevents the platform from effectively distinguishing between normal fluctuations and genuine anomalies when dealing with complex and ever-changing chemical environments, thus impacting production safety and efficiency. Therefore, there is an urgent need for a data processing method that can adapt to the production context to improve the intelligence level of anomaly detection. Summary of the Invention

[0004] The purpose of this invention is to provide a data processing method and system for an information management platform for chemical production, in order to solve the problems mentioned in the background section. Specific technical problems include how to dynamically select parameter configuration sets and associated data subsets based on production context labels, in order to address the issue of poor adaptability of anomaly detection algorithms in different production scenarios, leading to low detection accuracy.

[0005] To achieve the above objectives, one objective of this invention is a data processing method for an information management platform for chemical production, comprising the following steps:

[0006] S1. Acquire real-time data of chemical production, which includes system command sequences and equipment status signals; the system command sequences are command sequences issued by the central control system; the equipment status signals are real-time status feedback directly collected from field equipment.

[0007] This step, by acquiring the system command sequence issued by the central control system and the equipment status signals directly collected from field devices, constitutes a complete set of real-time data for chemical production. The role of this step is to provide an accurate and comprehensive data foundation for the entire method. Its effect is to ensure that the data source on which subsequent processing depends not only includes the production intention (i.e., the system command sequence) but also reflects the actual status of the equipment (i.e., equipment status signals). This lays a solid foundation for accurately understanding the current production situation and avoiding misjudgments caused by incomplete or missing data, thereby directly supporting the reliability of subsequent dynamic adaptation based on context.

[0008] S2. Based on the system command sequence and equipment status signals in the received real-time chemical production data, generate a production context label representing the current production situation. The specific process of generating the production context label includes:

[0009] The system instruction sequence is parsed to identify key instructions and their order, which is used to clarify the preset logical stages of production;

[0010] Simultaneously, aggregated analysis of device status signals is performed to monitor whether they match the expected status of the instruction sequence;

[0011] The parsed system instruction sequence is correlated and logically matched with the aggregated equipment status signals in real time. Judgments are made based on the rules defined in the pre-set production process knowledge base, and the judgment results are quantified into production context labels.

[0012] This step involves parsing, aggregating, and logically matching the system instruction sequence and equipment status signals with preset process rules to transform and quantify them into a production context label that accurately represents the current production situation. The effect of this step is to achieve intelligent perception and understanding of the dynamic production process, which elevates discrete, low-level data into contextual information with clear semantics. This solves the problem of fuzzy detection benchmarks caused by changes in contextual factors such as production stage and equipment health, and provides accurate and reliable judgment basis for dynamic configuration and correlation analysis based on production context labels in subsequent steps.

[0013] S3. Based on the production context label, dynamically select a corresponding parameter configuration set. The parameter configuration set is used to set the running parameters of the anomaly detection algorithm. The dynamic selection process of the parameter configuration set specifically includes:

[0014] Based on the production stage, equipment health status, and process execution status information encapsulated in the production context tags, the corresponding parameter configuration set is selected from a preset parameter configuration library by parsing the semantic information of the production context tags and performing matching reasoning based on preset process knowledge rules with the entries in the library.

[0015] Based on production context labels, a subset of data that is temporally associated with the production context labels is selected from real-time chemical production data. A correlation analysis is then performed on this selected subset. The correlation analysis process specifically includes:

[0016] Based on the timestamps corresponding to the production context labels and the duration of the production stage they represent, a fixed time window is defined from the real-time data stream of chemical production. All data points within this fixed time window that are relevant to the current production context are extracted to form a data subset. Correlation analysis is performed on this data subset, including time-series alignment and standardization of variables within the subset, and mining the dynamic coupling relationships between variables. Furthermore, correlation analysis determines whether there are abnormal interactions in the production process by calculating the cross-correlation between variables, and generates a comprehensive indicator. The specific process of generating the comprehensive indicator includes:

[0017] The cross-correlation results of variable pairs in the data subset are compared with a pre-set benchmark model, and the calculated deviation is quantified into a comprehensive index.

[0018] Step S3 directly addresses the technical issue of "how to dynamically select based on production context labels." Its function is reflected in two aspects: first, it dynamically selects the most suitable set of anomaly detection parameter configurations for the current context based on the labels, so that the algorithm's sensitivity, threshold, and other operating parameters can match the specific production stage (such as stable operation and drastic switching stages); second, it delineates the associated time window based on the labels and extracts data subsets for in-depth correlation analysis, generating a comprehensive index that reflects the dynamic coupling relationship between variables. The effect of this step is to realize the contextualized dynamic adjustment of analysis strategies and data range, effectively overcoming the rigidity of using fixed parameters and global data for analysis, and improving the pertinence of anomaly detection for specific production scenarios from both the perspectives of algorithm configuration and data focus.

[0019] S4. The anomaly detection algorithm, configured with the parameter set, is used to process the correlation analysis results and output a status indication signal. The specific processing steps include:

[0020] Anomaly detection algorithms are applied to comprehensive indicators obtained after correlation analysis. Anomaly scores are calculated using a machine learning-based outlier detection model, and these scores are compared with thresholds set in the parameter configuration set to determine whether the behavior deviates from the normal behavior pattern.

[0021] The process of generating status indication signals specifically includes:

[0022] The system maps anomaly scores to discrete state levels to generate an overall state rating and outputs anomaly confidence scores; it then performs anomaly tracing and classification to identify the variable combinations that trigger the anomalies and the anomaly types; finally, it encapsulates all elements into structured state indication signals.

[0023] Step S4 utilizes a context-configured anomaly detection algorithm to calculate the comprehensive indicators obtained from contextual analysis and generate structured status indication signals. Its purpose is to ultimately perform anomaly judgment and diagnosis. The effect is that, since the algorithm parameters and data foundation have been optimized according to the production context, the calculated anomaly scores, final status ratings, and source tracing information are more accurate and reliable. This solves the problem of low detection accuracy caused by the disconnect between the detection model and the current production context, and ultimately outputs a high-confidence status indication signal that can guide action.

[0024] The second objective of this invention is to provide a system for data processing in a chemical production information management platform, comprising a data acquisition module, a context label generation module, a dynamic configuration and data association module, and an anomaly detection and output module, wherein:

[0025] The data acquisition module acquires real-time data from chemical production.

[0026] The context label generation module generates a production context label that represents the current production situation based on the system command sequence and equipment status signals in the received real-time chemical production data.

[0027] The dynamic configuration and data association module dynamically selects a corresponding parameter configuration set based on the production context label. The parameter configuration set is used to set the running parameters of the anomaly detection algorithm.

[0028] The dynamic configuration and data association module selects a subset of data that is temporally associated with the production context label from real-time chemical production data based on the production context label, and performs correlation analysis on the selected data subset;

[0029] The anomaly detection and output module uses the anomaly detection algorithm configured with the parameter set to perform calculations on the results of the correlation analysis and outputs a status indication signal.

[0030] Compared with the prior art, the beneficial effects of the present invention are:

[0031] By introducing production context labels to dynamically perceive and quantify the current production situation, the anomaly detection process is no longer a rigid static analysis, but rather possesses adaptive and intelligent characteristics. The system can automatically match the optimal detection parameters and focus on related data based on contextual information such as the actual production stage and equipment health, improving the accuracy and timeliness of anomaly identification and effectively reducing false alarms and missed alarms caused by changes in operating conditions. At the same time, this method mines the deep coupling relationship between variables through correlation analysis and combines it with interpretable state indication signal output, which not only achieves accurate anomaly early warning, but also provides clear clues for anomaly tracing, thereby forming a closed-loop management for production safety and process optimization, and comprehensively enhancing the reliability and decision support capabilities of the chemical production information platform. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the overall method steps of the present invention;

[0033] Figure 2 This is a schematic diagram of the core process of step S3 of the present invention;

[0034] Figure 3 This is a schematic diagram of the overall module flow of the present invention.

[0035] In the diagram: 100, Data Acquisition Module; 200, Context Label Generation Module; 300, Dynamic Configuration and Data Association Module; 400, Anomaly Detection and Output Module. Detailed Implementation

[0036] To better understand the purpose, technical solution, and advantages of this application, the application is described and illustrated below in conjunction with the accompanying drawings and embodiments.

[0037] Unless otherwise defined, the technical or scientific terms used in this application shall have the general meaning understood by one of ordinary skill in the art to which this application pertains. Words such as “a,” “an,” “an,” “the,” “the,” and “these” used in this application do not indicate quantitative limitation and may be singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or modules (units) is not limited to the listed steps or modules (units) but may include steps or modules (units) not listed, or may include other steps or modules (units) inherent to these processes, methods, products, or devices. Words such as “connected,” “linked,” and “coupled” used in this application are not limited to physical or mechanical connections but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. Normally, the character " / " indicates that the objects before and after it are in an "or" relationship. The terms "first," "second," "third," etc., used in this application are merely to distinguish similar objects and do not represent a specific order of objects.

[0038] Next, please refer to Figure 1 One of the objectives of this embodiment is to provide a data processing method for a chemical production information management platform, comprising the following steps:

[0039] S1. Through a highly integrated data acquisition subsystem, real-time chemical production data is continuously acquired from various heterogeneous data sources at the chemical plant site. These heterogeneous data sources include, but are not limited to, distributed control systems, programmable logic controllers, sensor networks, monitoring and data acquisition systems, and production execution systems. They are responsible for different production stages, such as reactor temperature control, pump and valve operation monitoring, and material flow metering. The receiving process is first implemented through dedicated communication interface modules deployed at the data source end. These modules establish connections based on industrial network protocols (such as real-time transmission protocols using Ethernet technology) to ensure that data is continuously transmitted to the platform's data receiving server in a streaming manner. During transmission, the data is encapsulated into standard format messages, including timestamps, source identifiers, and data type identifiers, to facilitate subsequent parsing.

[0040] Specifically, the acquired real-time data from chemical production contains two types of key information: first, system command sequences, which are sequences of commands issued by the central control system or operator station, such as equipment start-up commands, parameter adjustment commands, and production stage switching commands. The system command sequences record the execution process of the control logic in an orderly manner; second, equipment status signals, which are real-time status feedbacks directly collected from field equipment, such as temperature sensor readings, pressure transmitter output values, motor operating status signals (such as running, stopped, and faulty), and valve position signals. Equipment status signals reflect the immediate operating conditions of the equipment.

[0041] S2. Based on the real-time chemical production data received from S1, a context analysis engine is used to generate production context labels representing the current production situation, specifically including:

[0042] First, the system command sequence in the real-time data of chemical production is analyzed to identify key commands such as "reactor heating program start", "material A feeding command", and "enter heat preservation stage command" and their sequence, thereby clarifying the preset logical stage of production; the analysis process specifically includes:

[0043] First, the command parsing module in the context analysis engine executes the process. Its core function is to decode and recognize the continuously incoming, timestamped raw command messages. The parsing module has a built-in key command dictionary, which pre-stores command keywords directly related to production process stages, such as "heating start," "feeding command," "heat preservation command," and "discharge command." It scans each arriving command message in real time, matching its content against the key command dictionary. Once a key command is identified, the parsing module extracts the command's precise timestamp, source identifier, and specific content, and temporarily stores it in a time-series command queue. Then, the logic stage identifier analyzes the sequence of instructions and checks whether the order of key instructions constitutes a valid production stage transition based on the preset process flow diagram (e.g., a complete reaction process must follow the sequence of "cleaning -> feeding -> heating -> holding -> cooling -> discharging"). For example, when the sequence of "reactor heating program start" and "enter holding stage command" is identified in sequence, the parsing process determines that the production logic stage has changed from "heating stage" to "holding stage". Finally, the process outputs a structured parsing result, clearly identifying the currently effective production logic stage and its start time.

[0044] Simultaneously, the system performs parallel aggregation analysis on equipment status signals from various field devices, monitoring whether signals such as actual reaction temperature, agitator current, feed valve opening, and pressure vessel level match the expected state of the command sequence. This aggregation analysis process is executed in parallel by the signal aggregation module in the context analysis engine. Its goal is to integrate and determine the consistency of heterogeneous, multi-frequency equipment status signals from different field devices, specifically including:

[0045] First, the aggregation module aligns all relevant signals within a short time window (such as the current production stage as analyzed in the previous step) based on the timestamps of the data points, forming a unified multivariate data snapshot. For example, after the system issues a "material A feeding instruction," the aggregation module continuously collects signals such as the opening degree of the feed valve, the reading of the material flow meter, and the liquid level signal of the reactor. Next, the module calls upon a pre-defined "desired state rule set" defined in the pre-built production process knowledge base for matching and monitoring. Each rule describes the conditions that relevant equipment status signals should meet under a fixed instruction. The aggregation analysis process compares the real-time aggregated signal snapshot with these rules, calculates the deviation between the actual and expected values, and determines whether the overall performance of each signal meets expectations. The output is a comprehensive assessment of the matching degree between equipment status and instruction expectations, including which signals are normal, which signals have deviations, and the degree of deviation.

[0046] Next, the parsed system instruction sequence is correlated and logically matched with the aggregated equipment status signals in real time. Judgments are made based on rules defined in a pre-defined production process knowledge base, and the judgment result is quantified into a machine-readable production context label, specifically including:

[0047] This process is the core of the scenario analysis engine's decision-making, completed by a rule inference engine. The rule inference engine receives the outputs of the first two processes: first, the parsed current production logic stage identifier; and second, a matching assessment report of the equipment status and instruction expectations. The inference engine then calls upon the highest-level association matching rules in the pre-set production process knowledge base. These rules are comprehensive clauses for judging the overall production scenario, in the form of "if certain conditions are met, then a certain scenario is inferred." For example, a rule is defined as: "If the current production logic stage is 'heat preservation stage,' and the actual reaction temperature is within the expected range, but the stirrer current remains above the normal threshold, then it is determined that the current production..." The following is an example: Production stage = 'Insulation stage', Equipment health status = 'Sub-healthy', Process execution status = 'Abnormal mechanical load'. The rule inference engine logically matches the received data with the preconditions of all rules in the knowledge base, searching for a complete or best-matching rule. The matching process may involve complex logical operations. Finally, the inference engine integrates the conclusions of the successfully matched rules, quantifying all judgment results, including the current stage, health status, and process status, into a structured, machine-readable data object, namely the final production context label. This label encapsulates all the core information of the current production situation, providing a decisive basis for subsequent adaptive anomaly detection.

[0048] The production context label, as a comprehensive identifier that encapsulates core information such as the current production stage, equipment health status, and process execution status, is output and passed to subsequent processing, providing a decisive basis for dynamically configuring anomaly detection parameters.

[0049] S3, please refer to Figure 2 Based on the production context label, a corresponding parameter configuration set is dynamically selected. This parameter configuration set is used to set the running parameters of the anomaly detection algorithm, specifically including:

[0050] Production context labels serve as machine-readable comprehensive identifiers of the current production context, encapsulating multi-dimensional information including production stage, equipment health status, and process execution status. A pre-defined parameter configuration library contains parameter configuration sets that match different production context labels. For example, when a production context label indicates the system is in the "reactor heating program start" stage, the context analysis engine automatically triggers the parameter configuration set call process for that fixed stage. This parameter configuration set contains a complete set of parameter definitions required for the anomaly detection algorithm to run, such as monitoring thresholds for upward trends in temperature variables, allowable fluctuation ranges, control limit coefficients in statistical process control, time window size, weighting factors for data smoothing, sensitivity parameters for mutation point detection, and correlation coefficient thresholds for multivariate correlation analysis.

[0051] The core of dynamic selection lies in real-time parsing of the semantic information carried by the production context tags and matching them with entries in the parameter configuration library. The matching process is not a simple keyword comparison, but rather reasoning based on pre-set process knowledge rules. For example, if the tags simultaneously display "entering the heat preservation stage" and "the agitator current is under high load," then a hybrid parameter set that takes into account both temperature stability monitoring and mechanical load anomaly detection is preferentially selected. This dynamic configuration mechanism ensures that the anomaly detection algorithm can adapt to different production scenarios. During the heating stage, parameters sensitive to the rate of temperature change are used, while during the feeding stage, the focus is on the coordinated anomalies of flow rate and pressure, thereby improving the accuracy and timeliness of detection and avoiding false alarms or missed alarms caused by fixed parameters.

[0052] Simultaneously, based on production context labels, a subset of data that is temporally associated with the production context labels is selected from real-time chemical production data, and correlation analysis is performed on the selected data subset, specifically including:

[0053] The process begins by defining a fixed time window from the continuously flowing real-time data stream of chemical production, based on the timestamp corresponding to the production context label and the duration of the production stage it represents. For example, if the production context label indicates the "Material A Feeding Instruction" execution stage, the system will extract all data points related to the feeding operation from the time the instruction is identified until the next key instruction (such as "Feeding End") appears, forming a logically coherent subset of data. This subset of data not only includes the timestamp of the instruction sequence itself, but also covers all relevant equipment status signals during this time period, such as the opening change curve of the feed valve, the cumulative reading of the material flow meter, the change of the liquid level in the reactor, and possibly related pressure and temperature auxiliary parameters.

[0054] After the data subset is selected, correlation analysis is initiated. The correlation analysis first performs time-series alignment and standardization on the variables within the data subset to eliminate biases caused by different sampling frequencies; specifically, this includes:

[0055] A unified time axis is defined for the data subset, with the minimum interval determined by the variable with the highest sampling frequency within the subset. For variables with lower sampling frequencies, a linear interpolation algorithm is used to calculate approximate values ​​for each time point on the unified time axis based on the data values ​​before and after them. This ensures that all variables have corresponding data values ​​at each unified time point, forming a strictly aligned multivariate time series. Standardization is then performed. For each variable, its arithmetic mean and standard deviation are calculated over the entire data subset's time window. Each data value of the variable is then subtracted from the mean and divided by the standard deviation, transforming variables with different dimensions and numerical ranges into a dimensionless numerical sequence with a mean of zero and a standard deviation of one. After preprocessing, dynamic coupling relationships are explored. Cross-correlation functions are used to calculate the Pearson correlation coefficient at multiple time offsets for process-related variable pairs, such as feed valve opening and material flow rate. This determines the most significant time delay and correlation strength at which a change in one variable causes a change in another, thus quantitatively describing the dynamic lead-follow relationship between variables.

[0056] The analysis then focuses on uncovering the dynamic coupling relationships between variables. For example, it examines whether the following characteristics between the feed valve opening command and the actual flow rate reading are normal, and checks whether the reactor temperature and agitator current exhibit the expected coordinated change pattern during the feeding process. By calculating the cross-correlation between variables, it determines whether there are abnormal interactions during production and generates comprehensive indicators. Specifically, this includes:

[0057] From a pre-built production process knowledge base, a standard cross-correlation function calculated under normal historical operating conditions for a specific pair of variables (such as pump start / stop commands and outlet pressure) is retrieved as a benchmark model. This benchmark model includes the maximum correlation coefficient value that should exist under normal conditions and the time delay value corresponding to reaching that maximum correlation coefficient. In real-time analysis, the same method is used to calculate the cross-correlation function for the same pair of variables in the real-time data subset, obtaining the real-time maximum correlation coefficient and time delay. Then, the real-time calculation results are compared with the benchmark model: the relative attenuation percentage of the real-time maximum correlation coefficient and the benchmark maximum correlation coefficient are calculated, and the absolute difference between the real-time time delay and the benchmark time delay is calculated. Finally, these deviations are quantified into a comprehensive index, for example, generating a comprehensive index called "Control Loop Performance Index." This index is calculated by a weighted formula, where the higher-weighted real-time correlation coefficient attenuation percentage and the lower-weighted time delay difference are combined. The larger the index value, the more severe the abnormal interaction.

[0058] The comprehensive indicators obtained from this correlation analysis can reveal deeper issues behind the surface data, such as potential equipment performance degradation, control loop failure, or process deviations, providing support for accurate anomaly diagnosis.

[0059] S4. The anomaly detection algorithm, configured with the specified parameters, is used to process the correlation analysis results and output a status indication signal, specifically including:

[0060] The dynamically configured anomaly detection algorithm is applied to the comprehensive index obtained from the correlation analysis. Specifically, the anomaly detection algorithm first initializes its own operating parameters based on the parameter configuration set dynamically selected in step S3. These parameters are adapted to the fixed operating conditions defined by the current production context label. For example, in the "reactor heat preservation stage", a monitoring threshold with extremely high requirements for temperature stability and a small allowable fluctuation range are adopted. Meanwhile, for correlated variables such as stirrer current, a detection logic that focuses on trend stability rather than absolute value is enabled.

[0061] The input object of the operation is not the original, isolated equipment status signal, but a comprehensive index obtained after correlation analysis. This comprehensive index reveals the dynamic coupling relationship between key process parameters, such as the following error between the feed valve opening and the flow reading, and the residual of the coordinated change between reaction temperature and stirring power.

[0062] An outlier detection model based on machine learning performs joint analysis of comprehensive indicators to determine whether they deviate from the normal behavior pattern defined by a parameter configuration set; specifically including:

[0063] First, a set of comprehensive indicators generated by correlation analysis is combined into a multi-dimensional feature vector. The outlier detection model used is the Isolation Forest algorithm, which learns the spatial distribution density characteristics of normal comprehensive indicator feature vectors during the training phase with historical normal data. The real-time arriving comprehensive indicator feature vectors are input into the constructed isolation forest, and the average path length required for the vector to be isolated is calculated. This length is standardized and converted into an anomaly score between zero and one. Then, the calculated anomaly score is compared with an anomaly score threshold set in the parameter configuration set: if the anomaly score is greater than the threshold, the current comprehensive indicator feature vector is determined to be an outlier, i.e., deviating from the normal behavior pattern; if the anomaly score is less than or equal to the threshold, it is determined to be normal. This threshold-based binary judgment, along with the specific value of the anomaly score, serves as the core basis for generating a status indicator signal. The process of generating the status indicator signal specifically includes:

[0064] The anomaly score output by the anomaly detection algorithm is parsed and mapped. This anomaly score, generated by a machine learning-based outlier detection model, quantifies the degree to which the current comprehensive index deviates from the normal behavior pattern. A multi-level rule base is preset, which maps continuous anomaly score intervals to discrete state levels. For example, when the anomaly score is below a first threshold, it is mapped to a "normal" level; when the anomaly score is between the first and second thresholds, it is mapped to a "warning" level; and when the anomaly score is above the second threshold, it is mapped to an "abnormal" level. This mapping result constitutes the "overall state rating" in the state indication signal. At the same time, the anomaly score itself, after normalization, is directly output as the "anomaly confidence score".

[0065] Next, anomaly tracing and classification are performed. The contribution of each dimension (i.e., the correlation index of each variable pair) constituting the comprehensive index feature vector to the overall anomaly score is analyzed, thereby identifying the one or more variable combinations with the highest contribution and marking them as the "main variable or variable combination identifier causing the anomaly." Furthermore, based on a pre-built anomaly pattern knowledge base, the performance patterns of the anomaly feature vectors are matched. If the feature vector shows a slow weakening of the coupling relationship between key process parameters, it is classified as "gradual performance degradation"; if it shows a sudden break in the relationship, it is classified as "sudden equipment failure." Finally, all the above elements—overall status rating, anomaly confidence score, main variable identifier, and preliminary anomaly type classification—are encapsulated into a structured, machine-readable data object. This data object is the final generated status indication signal that can be directly parsed and utilized by the upper-level system.

[0066] As can be seen from the above description, the data processing method for a chemical production information management platform provided in this embodiment has the following technical effects:

[0067] The data processing method of the chemical production information management platform provided in this embodiment achieves precise adaptive matching between anomaly detection parameter configuration and the scope of associated data analysis by constructing context labels for dynamically perceiving production situations. This effectively overcomes the fundamental defect of poor adaptability of traditional fixed parameter models under complex and variable working conditions. The method deeply integrates instruction logic parsing, equipment status aggregation, and process rule reasoning to ensure the accuracy of production situation judgment. This drives the anomaly detection algorithm to focus on key data subsets and conduct in-depth correlation analysis, ultimately generating structured indication signals that combine status rating, confidence level, and traceability information. This reduces false alarms and missed alarms at the source, significantly improves the accuracy and timeliness of anomaly detection, and forms a closed-loop management system from situation perception to intelligent diagnosis, providing reliable protection for production safety and process optimization.

[0068] Please see Figure 3The second objective of this embodiment is to provide a system for data processing of a chemical production information management platform, including a data acquisition module 100, a context label generation module 200, a dynamic configuration and data association module 300, and an anomaly detection and output module 400, wherein:

[0069] Data acquisition module 100 acquires real-time data from chemical production;

[0070] The context label generation module 200 generates a production context label representing the current production situation based on the system instruction sequence and equipment status signals in the received real-time chemical production data.

[0071] The dynamic configuration and data association module 300 dynamically selects a corresponding parameter configuration set based on the production context label. The parameter configuration set is used to set the running parameters of the anomaly detection algorithm.

[0072] The dynamic configuration and data association module 300 selects a subset of data that is temporally associated with the production context label from the real-time data of chemical production based on the production context label, and performs correlation analysis on the selected data subset;

[0073] The anomaly detection and output module 400 uses the anomaly detection algorithm configured with the parameter set to perform calculations on the results of the correlation analysis and outputs a status indication signal.

[0074] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A data processing method for an information management platform for chemical production, characterized in that, The methods and steps include the following: S1. Obtain real-time data on chemical production; S2. Based on the system instruction sequence and equipment status signals in the received real-time chemical production data, generate a production context label that represents the current production situation; The process of generating the production context label specifically includes: The system instruction sequence is parsed to identify key instructions and their order, which is used to clarify the preset logical stages of production; Simultaneously, the device status signals are aggregated and analyzed to monitor whether they match the expected state of the instruction sequence; The parsed system instruction sequence is correlated and logically matched with the aggregated equipment status signals in real time, and the judgment is made according to the rules defined in the preset production process knowledge base. The judgment result is quantified into the production context label. S3. Based on the production context label, dynamically select a corresponding parameter configuration set, which is used to set the running parameters of the anomaly detection algorithm; Based on the production context label, a subset of data that is temporally associated with the production context label is selected from the real-time chemical production data, and a correlation analysis is performed on the selected data subset; the correlation analysis process specifically includes: Based on the timestamps corresponding to the production context labels and the duration of the production stage they represent, a fixed time window is defined from the real-time data stream of chemical production. All data points within this fixed time window that are relevant to the current production context are extracted to form a data subset. A correlation analysis is performed on this data subset, including time-series alignment and standardization of variables within the subset, and mining the dynamic coupling relationships between variables. The correlation analysis determines whether there are abnormal interactions during the production process by calculating the cross-correlation between variables, and generates a comprehensive index. The specific process of generating the comprehensive index includes: The cross-correlation results of variable pairs in the data subset are compared with a preset benchmark model, and the calculated deviation is quantified into the comprehensive index. S4. The anomaly detection algorithm configured with the parameter set is used to perform calculations on the results of the correlation analysis and output a status indication signal.

2. The data processing method of the chemical production information management platform according to claim 1, characterized in that, The real-time data for chemical production includes system command sequences and equipment status signals; the system command sequences are command sequences issued by the central control system; and the equipment status signals are real-time status feedback directly collected from field equipment.

3. The data processing method of the chemical production information management platform according to claim 1, characterized in that, The dynamic selection process of the parameter configuration set specifically includes: Based on the production stage, equipment health status, and process execution status information encapsulated in the production context tag, a corresponding parameter configuration set is selected from a preset parameter configuration library by parsing the semantic information of the production context tag and performing matching reasoning based on preset process knowledge rules with the entries in the library.

4. The data processing method of the chemical production information management platform according to claim 1, characterized in that, The calculation process in step S4 specifically includes: The anomaly detection algorithm is applied to the comprehensive index obtained after correlation analysis. An anomaly score is calculated using a machine learning-based outlier detection model, and the anomaly score is compared with a threshold set in the parameter configuration set to determine whether it deviates from the normal behavior pattern.

5. The data processing method of the chemical production information management platform according to claim 1, characterized in that, The process of generating the status indication signal specifically includes: The system maps anomaly scores to discrete state levels to generate an overall state rating and outputs anomaly confidence scores; it then performs anomaly tracing and classification to identify the variable combinations that trigger the anomalies and the anomaly types; finally, it encapsulates all elements into structured state indication signals.

6. A system using a data processing method comprising the chemical production information management platform according to any one of claims 1-5, characterized in that, It includes a data acquisition module (100), a context label generation module (200), a dynamic configuration and data association module (300), and an anomaly detection and output module (400), wherein: The data acquisition module (100) acquires real-time data of chemical production; The context label generation module (200) generates a production context label representing the current production situation based on the system instruction sequence and equipment status signal in the received real-time chemical production data; The dynamic configuration and data association module (300) dynamically selects a corresponding parameter configuration set based on the production context label. The parameter configuration set is used to set the running parameters of the anomaly detection algorithm. The dynamic configuration and data association module (300) selects a subset of data that is temporally associated with the production context label from the real-time chemical production data based on the production context label, and performs correlation analysis on the selected data subset; The anomaly detection and output module (400) uses the anomaly detection algorithm set by the parameter configuration set to calculate the result of the correlation analysis and outputs a status indication signal.

Images