Steel product carbon label data linkage monitoring and abnormal traceability method and system with production equipment operation parameters

Abstract

This application relates to the field of process control and data fusion technology in the steel industry, and discloses a method and system for linked monitoring and anomaly tracing of carbon label data of steel products and operating parameters of production equipment. The method involves deploying sensor arrays to collect equipment operating parameters in real time and extracting carbon label data from a carbon footprint accounting system; constructing a spatiotemporal alignment mapping model, utilizing time tolerance parameters and material flow tracking to solve the problem of spatiotemporal misalignment of multi-source heterogeneous data; employing dual-threshold joint judgment logic to detect carbon label-equipment correlation anomalies; accurately tracing and locating the root cause equipment based on anomaly contribution, and generating equipment control commands after fault feature matching; and issuing commands through an industrial control network to achieve equipment status adjustment and closed-loop verification. This application achieves deep integration and closed-loop feedback of carbon emission data and physical equipment operating conditions, solving the technical problem of existing independent carbon accounting and equipment control, and improving the level of refined carbon emission management and anomaly response efficiency.

Description

Technical Field

[0001] This application relates to the field of carbon emission management and industrial automation control technology in steel production processes, and in particular to a technology for linking steel product carbon label data with production equipment operating parameters for monitoring and anomaly tracing. Background Technology

[0002] In actual production, steel companies typically deploy carbon footprint accounting systems to calculate the carbon emissions of their products and use carbon labels to identify and trace the carbon emission data of each production batch. Meanwhile, key equipment in the production process, such as heating furnaces, rolling mills, and finishing equipment, are equipped with corresponding data acquisition and monitoring systems to monitor the operating status parameters of the equipment in real time.

[0003] Taking a hot-rolled steel production line as an example, when the carbon label data for a certain coil of steel shows an abnormally high level, production managers need to determine whether the anomaly is caused by fluctuations in raw material input, energy metering errors, or abnormal equipment operating conditions. Under the traditional technical architecture, the carbon footprint accounting system and the equipment monitoring system belong to different levels of enterprise informatization. The former calculates carbon emissions based on batch order data from the manufacturing execution system, while the latter monitors equipment status based on sensor data collected by programmable logic controllers (PLCs). The two systems operate independently and their data is not shared. When carbon labels show anomalies, technicians often need to manually retrieve equipment operating data for the relevant time period and check for possible abnormal equipment one by one. This process is not only time-consuming and labor-intensive, but also, due to the lack of systematic data correlation analysis methods, it is difficult to accurately pinpoint the root cause of the anomaly.

[0004] In another typical scenario, when a decrease in the combustion efficiency of a heating furnace leads to an increase in gas consumption per unit of product, this change in equipment operating conditions is directly reflected in an increase in carbon emission data. However, because the timestamps of carbon label data are based on the completion time of production batches, while the timestamps of equipment parameters are based on the real-time acquisition cycle of sensors, there is an inherent offset between the two types of data in the time dimension. At the same time, since the same production process may be completed by multiple parallel devices, there is a complex mapping relationship between process codes and equipment codes, making it difficult to establish a precise correspondence between the two types of data in the spatial dimension. This spatiotemporal misalignment problem of multi-source heterogeneous data makes it difficult for existing technologies to effectively integrate carbon label data and equipment operating parameters.

[0005] Furthermore, even if carbon tag anomalies can be detected and initially linked to certain devices, current technology lacks a closed-loop mechanism from anomaly detection to device control. The current approach typically involves the system generating an alarm after detecting an anomaly, followed by manual assessment of the cause and issuance of device adjustment commands. This reliance on manual intervention results in a long anomaly response cycle, making it difficult to achieve real-time intervention and rapid correction of carbon emission anomalies.

[0006] Therefore, there is an urgent need for a new technical solution that can achieve spatiotemporal alignment and fusion of carbon tag data and equipment operating parameters, establish a correlation detection mechanism between carbon emission anomalies and equipment operating condition anomalies, achieve precise traceability and location at the equipment level, and form an automated closed loop from anomaly detection to equipment control, thereby effectively improving the level of precision in carbon emission management of steel products. Summary of the Invention

[0007] The purpose of this application is to provide a method and system for linking carbon label data of steel products with operating parameters of production equipment and for tracing anomalies, so as to solve the problems mentioned in the background art.

[0008] This application discloses a method for linking carbon label data of steel products with operating parameters of production equipment for monitoring and anomaly tracing, including the following steps: S1: Through sensor arrays deployed on key equipment in steel production processes, physical parameters characterizing the operating conditions of the equipment are collected in real time and transmitted to the edge computing gateway via the industrial network; S2: Obtain real-time carbon label data of the current production batch and its associated process list information from the carbon footprint accounting system, and determine the target equipment set based on the mapping relationship between process codes and equipment codes; S3: Construct a spatiotemporal alignment mapping model between the real-time carbon tag data and the physical parameters; extract the equipment parameter sequence of each device in the target equipment set within the corresponding time window from the equipment operation parameter time series database; extract statistical features from the equipment parameter sequence to generate an equipment operation feature vector; and concatenate it with the real-time carbon tag data to form a fusion feature record. S4: Based on the fusion feature record, calculate the carbon tag deviation index and the equipment parameter anomaly index. Only when both exceed their respective preset thresholds is it determined that the carbon tag-equipment association is abnormal. S5: In response to the carbon tag-equipment association anomaly, calculate the anomaly contribution of each device to identify the root cause device, and match the parameter anomaly pattern of the root cause device with a preset fault feature library to determine the anomaly cause type; S6: Match a control scheme from the preset control strategy library according to the type of abnormal cause, generate equipment control instructions and send them to the controller of the root cause device for execution; S7: Monitor the carbon label deviation index and the abnormal equipment parameter index of subsequent production batches. When the index of a preset number of consecutive batches returns to normal, the abnormality is determined to be eliminated, and the preset threshold is updated according to the abnormality handling result.

[0009] In a preferred embodiment, the construction of the spatiotemporal alignment mapping model in S3 further includes: The corresponding time window is determined based on the start and end timestamps of the current production batch, combined with the time tolerance parameter determined by the process type. Given the one-to-many mapping between process codes and equipment codes in the process list information, the target equipment that actually performs the operation is located in the target equipment set based on the material flow tracking information in the production process.

[0010] In a preferred embodiment, in step S1, the sensor group employs an adaptive acquisition frequency strategy, specifically including: Low-frequency data acquisition mode is used during steady-state operation of the equipment; The rate of change of the physical parameters is monitored in real time, and when the rate of change exceeds a preset threshold, the system automatically switches to high-frequency acquisition mode. When the rate of change recovers to below the preset change threshold and remains below it for a preset delay time, the system switches back to the low-frequency acquisition mode. The device parameter sequence in S3 consists of data collected according to the adaptive acquisition frequency strategy.

[0011] In a preferred embodiment, in S1, the physical parameters include at least two of the following: furnace temperature, gas flow rate and oxygen content in the flue gas of the heating furnace, main motor current, rolling force and roll speed of the rolling mill, and shearing force and straightening roll pressure of the finishing equipment.

[0012] In a preferred embodiment, step S2, before determining the target equipment set, further includes a pre-launch / unlaunched furnace verification step: Check whether there are any missing list data for a specific process in the process list information; If any data is missing, the current production batch is determined to be in a state of being ready for launch but not yet launched, marked as invalid data, and subsequent steps are terminated. The "ready-to-launch but not yet launched" status refers to a state in which a production order has been issued in the manufacturing management system, but no actual consumption data has been generated on the subsequent production equipment.

[0013] In a preferred embodiment, step S3 involves extracting the device parameter sequence of each device in the target device set within a corresponding time window from the device operating parameter time series database. Specifically, this includes: Obtain the start timestamp of the current production batch. and end timestamp ; The time tolerance parameter is determined according to the process type. For continuous processes, a first tolerance value is used, and for batch processes, a second tolerance value is used, wherein the second tolerance value is greater than the first tolerance value. Using the start timestamp minus the time tolerance parameter as the starting point of the window, and the end timestamp plus the time tolerance parameter as the ending point of the window, a time interval is extracted from the device operating parameter time series database. The sequence of device parameters within.

[0014] In a preferred embodiment, in step S3, extracting statistical features from the device parameter sequence to generate a device operating feature vector includes: The device parameter sequence within the time window is calculated, and at least three statistical features are extracted, including the mean, standard deviation, maximum value, minimum value, and energy consumption integral value. The extracted statistical features are combined to form the device operation feature vector.

[0015] In a preferred embodiment, in step S4, the carbon label deviation index is calculated using the following formula: in, This refers to the carbon label deviation index. This represents the carbon emissions value for the current production batch. This is the historical benchmark value for similar products. This indicates the absolute value operation; The abnormal equipment parameter index is calculated using the Mahalanobis distance between the equipment's operating feature vector and the equipment's historical normal operating range. To characterize.

[0016] In a preferred embodiment, the preset threshold in S4 is determined using a dynamic threshold generation method, with the following formula: in, The preset threshold, The average of historical data. The standard deviation of historical data, Confidence coefficient; The preset threshold is updated on a rolling basis according to a preset period based on the latest historical data.

[0017] In a preferred embodiment, in step S4, if only one condition is met—either the carbon label deviation index exceeds its corresponding preset threshold or the equipment parameter abnormality index exceeds its corresponding preset threshold—it is determined to be a suspected abnormality and marked for review.

[0018] In a preferred embodiment, in step S5, the calculation of the abnormal contribution of each device in the target device set is performed using the following formula: in, For the first in the target device set Abnormal contribution of each device For the first Carbon emission weighting factor for each equipment corresponding to a process. For the first Abnormal indicators of equipment parameters for each device. This is the sum of abnormal device parameter indicators for all devices in the target device set. The root cause identification device includes: devices whose cumulative contribution value exceeds a preset contribution threshold and are arranged in descending order of the abnormal contribution degree as the root cause device.

[0019] In a preferred embodiment, the preset fault feature library in S5 is constructed in the following manner: Collect time-series data of device parameters from historical anomaly cases; Features are extracted from the time-series data, and anomaly type labels marked by process experts are obtained; A fault mode-feature mapping model is trained using a machine learning classification algorithm to form the pre-set fault feature library.

[0020] In a preferred embodiment, the equipment control command generated in S6 includes at least one of the following: furnace gas valve opening adjustment amount, mill speed correction value, and cooling water flow rate setting value.

[0021] In a preferred embodiment, S6 further includes a security verification step: Before issuing the equipment control command through the industrial control network, read the safe operating range parameters of the root cause equipment; Verify whether the specific adjustment parameters in the device control command are within the safe operating range; If the specific adjustment parameter exceeds the safe operating range, the specific adjustment parameter will be limited or the issuance will be terminated and a manual confirmation request will be generated.

[0022] In a preferred embodiment, S6 further includes performing a confirmation step: After the device control command is issued, the execution feedback signal returned by the controller of the root cause device is received; The execution feedback signal confirms that the device control command has taken effect.

[0023] In a preferred embodiment, the closed-loop verification and parameter optimization steps in S7 specifically include: Based on the convergence of the carbon label deviation index after anomaly handling, the preset threshold in S4 is adjusted using an online learning method. The system receives execution feedback from the controller regarding the device control commands, and optimizes the strategy parameters of the preset control strategy library in step S6 based on the execution feedback.

[0024] This application also proposes a system for linking carbon label data of steel products with operating parameters of production equipment for monitoring and anomaly tracing, including: The equipment parameter acquisition module is used to collect physical parameters that characterize the operating conditions of the equipment in real time through a sensor group deployed on key equipment in steel production processes, and transmit them to the edge computing gateway through the industrial network. The carbon label association parameter extraction module is used to obtain real-time carbon label data of the current production batch and its associated process list information from the carbon footprint accounting system, and determine the target equipment set based on the mapping relationship between process code and equipment code; The spatiotemporal alignment mapping module is used to construct a spatiotemporal alignment mapping model between the real-time carbon tag data and the physical parameters. It extracts the equipment parameter sequence within the time window based on the start and end timestamps of the production batch and the time tolerance parameter determined by the process type. It processes the one-to-many mapping relationship between the process and the equipment based on the material flow tracking information and generates a fused feature record. The anomaly joint detection module is used to calculate the carbon tag deviation index and the equipment parameter anomaly index based on the fused feature record, and to determine the carbon tag-equipment association anomaly only when both exceed their respective preset thresholds. An anomaly tracing and localization module is used to calculate the anomaly contribution of each device to identify the root cause device when the carbon tag-equipment association anomaly is detected, and to match the parameter anomaly pattern of the root cause device with a preset fault feature library to determine the anomaly cause type. The linkage control module is used to match a control scheme from a preset control strategy library according to the type of abnormal cause, generate equipment control instructions, and send the equipment control instructions to the controller of the root cause device for execution; The closed-loop verification module is used to monitor the carbon label deviation index and equipment parameter anomaly index of subsequent batches to verify the anomaly handling effect, and update the preset threshold and the strategy parameters of the preset control strategy library according to the anomaly handling results.

[0025] This application also proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for linking carbon label data of steel products with operating parameters of production equipment and for anomaly tracing.

[0026] The method and system for linking steel product carbon label data with production equipment operating parameters for monitoring and anomaly tracing provided in this application address the technical problems in existing technologies, such as the independence of carbon label accounting systems and equipment control systems, difficulties in spatiotemporal alignment of multi-source heterogeneous data, and the lack of a closed-loop mechanism from anomaly detection to equipment control. Through the organic combination of multiple technical means, significant technical effects have been achieved.

[0027] Regarding real-time acquisition of equipment operating parameters, this application utilizes a sensor array deployed on key steel production equipment to collect physical parameters characterizing the equipment's operating conditions in real time. These physical parameters are then transmitted to an edge computing gateway via an industrial network, providing a physical-level data foundation for subsequent spatiotemporal alignment mapping. Furthermore, the sensor array employs an adaptive acquisition frequency strategy. During steady-state operation, it uses a low-frequency acquisition mode. When the rate of change of physical parameters exceeds a preset threshold, it automatically switches to a high-frequency acquisition mode. Once the rate of change recovers and remains for a preset delay, it switches back to the low-frequency acquisition mode. This adaptive acquisition frequency strategy enables high-density sampling data to capture abnormal transient characteristics during periods of equipment anomalies, while effectively reducing sensor load and network bandwidth consumption during steady-state periods, achieving a good balance between data acquisition accuracy and system resource consumption.

[0028] Regarding the extraction of carbon label-related parameters, this application obtains real-time carbon label data of the current production batch and its associated process list information from the carbon footprint accounting system. Based on the mapping relationship between process codes and equipment codes, it determines the target equipment set, clarifying the data range for subsequent analysis. Furthermore, before determining the target equipment set, this application introduces a pre-launch / pre-fired furnace verification step. This step identifies invalid batch data in the pre-launch / pre-fired furnace state by checking for missing data in the process list information for specific processes. The pre-launch / pre-fired furnace state refers to an abnormal state where a production order has been issued in the manufacturing management system, but no actual consumption data has been generated on the subsequent production equipment. Through this verification step, the system can effectively exclude batches with incomplete data, avoiding incorrect anomaly judgments based on incomplete data, thereby effectively shortening the verification range and improving overall analysis efficiency.

[0029] Regarding spatiotemporal alignment mapping, this application constructs a spatiotemporal alignment mapping model between real-time carbon tag data and physical parameters, which is a core technical means to solve the problem of multi-source heterogeneous data fusion. In the time dimension, the system extracts the equipment parameter sequence within a time window from the equipment operation parameter time series database based on the start and end timestamps of the current production batch, combined with a time tolerance parameter determined by the process type. A smaller first tolerance value is used for continuous processes, and a larger second tolerance value is used for batch processes, thus accommodating timestamp offsets between the two systems. In the spatial dimension, for the one-to-many mapping between process codes and equipment codes, the system locks the target equipment actually performing the operation based on material flow tracking information during the production process, eliminating interference from irrelevant equipment. In terms of feature fusion, the system extracts statistical features, including mean, standard deviation, maximum, minimum, and energy consumption integral values, from the equipment parameter sequence to generate an equipment operation feature vector, and concatenates it with real-time carbon tag data to form a fused feature record. Through the above-mentioned spatiotemporal alignment mapping steps, this application overcomes the technical bias of existing technologies where carbon label data and equipment physical status are disconnected, and achieves deep integration of business-level carbon emission data and physical-level equipment operating condition data, laying a data foundation for subsequent joint anomaly detection.

[0030] In terms of joint anomaly detection, this application calculates carbon label deviation indicators characterizing the degree of carbon emission anomalies and equipment parameter anomaly indicators characterizing the degree of equipment operating condition anomalies based on fused feature records. A carbon label-equipment correlation anomaly is only determined when both indicators exceed their respective preset thresholds. This dual-indicator joint determination mechanism is fundamentally different from existing methods based on single-dimensional data for anomaly detection. It effectively reduces the false alarm rate caused by fluctuations in carbon label data or occasional anomalies in equipment parameters, improving the accuracy and reliability of anomaly determination. Furthermore, when only a single condition is met, it is determined as a suspected anomaly and marked for review, avoiding the omission of potential anomalies and providing a basis for subsequent manual review. In addition, this application uses a dynamic threshold generation method to determine the preset threshold and updates it on a rolling basis according to a preset period based on the latest historical data. This allows the detection threshold to adapt to gradual adjustments in production processes and baseline drift caused by equipment aging, exhibiting better robustness compared to fixed thresholds.

[0031] In terms of tracing and locating abnormal equipment, this application responds to detected carbon tag-equipment-related anomalies by calculating the anomaly contribution of each device in the target equipment set based on anomaly indicators of equipment parameters. A carbon emission weighting factor is introduced to comprehensively consider the degree of anomaly and the weighting percentage of carbon emissions for each device. Devices are sorted in descending order of anomaly contribution and those with cumulative contribution values ​​exceeding a preset contribution threshold are identified as root cause devices. This identification method focuses on the few key devices that contribute the most to the anomaly, avoiding indiscriminate control of all devices. Furthermore, the system matches the parameter anomaly patterns of the root cause devices with a pre-set fault feature library to determine the type of anomaly. This fault feature library is generated by collecting historical anomaly cases, extracting features, labeling anomaly types with process experts, and training with machine learning classification algorithms, covering common equipment anomaly patterns. Through the above anomaly tracing and locating steps, this application achieves a leap from anomaly detection to root cause location, providing clear control targets and targeted control basis for subsequent coordinated control.

[0032] Regarding the generation and execution of linkage control commands, this application matches control schemes from a pre-set control strategy library based on the type of abnormal cause, generates equipment control commands containing specific adjustment parameters, and sends these commands to the controller of the root cause device for execution via an industrial control network. The equipment control commands include adjustments to the opening of the heating furnace gas valve, mill speed correction values, and cooling water flow rate setpoints, enabling targeted control measures for different types of abnormal causes. Furthermore, this application performs a safety verification step before issuing the command to check whether the specific adjustment parameters in the equipment control command are within the safe operating range of the root cause device. If they exceed this range, a limiting process is performed or a manual confirmation request is generated, effectively preventing equipment damage or safety accidents caused by excessive control. In addition, this application performs a confirmation step after issuing the command, receiving execution feedback signals from the controller to confirm that the equipment control command has taken effect, ensuring reliable execution of the control command. Through the above-mentioned linkage control steps, this application forms a complete closed loop from anomaly detection to equipment control, extending carbon tag accounting from a post-event statistical function to an active management function that can provide real-time feedback and control of the production process, and significantly shortening the anomaly response time from hours in the traditional manual processing mode to minutes in the automatic processing mode.

[0033] Regarding closed-loop verification and parameter optimization, this application continuously monitors the carbon label deviation indicators and abnormal equipment parameter indicators of subsequent production batches after the equipment control command is executed. When the indicators of a preset number of consecutive batches recover to the normal range, the anomaly is determined to be eliminated. This determination method based on the recovery of indicators from multiple consecutive batches is more robust than that based on a single batch, avoiding misjudgments caused by accidental fluctuations. Furthermore, the system adjusts preset thresholds using an online learning method based on the anomaly handling results and optimizes the strategy parameters of the preset control strategy library based on the controller's execution feedback. This allows the system to continuously learn and improve, continuously enhancing the accuracy and efficiency of carbon label anomaly detection and processing, and achieving adaptive evolution of the system.

[0034] In summary, this application constructs a complete technical system for the linkage monitoring and anomaly tracing of carbon tag data and production equipment operating parameters in the steel industry through the organic combination and synergistic cooperation of technologies such as real-time acquisition of equipment operating parameters, extraction of carbon tag-related parameters, spatiotemporal alignment mapping, joint anomaly detection, anomaly equipment tracing and location, generation and execution of linkage control commands, and closed-loop verification and parameter optimization. This technical system overcomes the limitations of the existing technology's independent technical architecture between the carbon tag accounting system and the equipment control system, achieving precise correlation between carbon tag anomalies and equipment physical status, precise equipment-level tracing and location, and an automated closed loop from anomaly detection to equipment control. This provides effective technical support for the refined management of carbon emissions in the steel production process.

[0035] The specification of this application contains numerous technical features distributed across various technical solutions. Listing all possible combinations of these technical features (i.e., technical solutions) would make the specification excessively lengthy. To avoid this problem, the various technical features disclosed in the above-described invention, the various technical features disclosed in the following embodiments and examples, and the various technical features disclosed in the accompanying drawings can be freely combined to form various new technical solutions (all of which are considered to have been described in this specification), unless such a combination of technical features is technically infeasible. For example, one example discloses feature A+B+C, and another example discloses feature A+B+D+E. Features C and D are equivalent technical means that serve the same function, and technically only one needs to be used; they cannot be used simultaneously. Feature E can technically be combined with feature C. Therefore, the solution A+B+C+D should not be considered as described because it is technically infeasible, while the solution A+B+C+E should be considered as described. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating the method for linking carbon label data of steel products with operating parameters of production equipment and tracing anomalies according to the first embodiment of this application.

[0037] Figure 2 This is a schematic diagram of the structure of the steel product carbon label data and production equipment operation parameter linkage monitoring and anomaly tracing system according to the second embodiment of this application.

[0038] Figure 3 This is a detailed schematic diagram of spatiotemporal alignment mapping. Detailed Implementation

[0039] In the following description, many technical details are presented to help the reader better understand this application. However, those skilled in the art will understand that the technical solutions claimed in this application can be implemented even without these technical details and various variations and modifications based on the following embodiments.

[0040] Explanation of some concepts: Real-time carbon labeling data refers to the carbon footprint data identifier assigned to a batch of steel products (such as steel coils and billets) immediately after production, based on the actual raw materials, energy media, and auxiliary materials consumed during the production process, calculated by a carbon footprint accounting system embedded in the production process. This label dynamically reflects carbon emissions under specific production conditions and possesses traceability, timeliness, and uniqueness. Its data content includes product code, process code, raw material consumption, energy consumption, and calculated carbon emissions.

[0041] Physical parameters refer to process parameters that characterize the operating conditions of equipment, collected in real time by sensor arrays deployed on key equipment in steel production processes. Depending on the type of equipment, physical parameters include, but are not limited to: furnace temperature, gas flow rate, and oxygen content in flue gas of heating furnaces; main motor current, rolling force, and roll speed of rolling mills; and shearing force and straightening roll pressure of finishing equipment.

[0042] The target equipment set refers to the set of all equipment involved in the production of the current batch of products, determined by the mapping relationship between process codes and preset equipment codes, based on the process list information of the current production batch. It is the basis for spatiotemporal alignment mapping and anomaly tracing and location.

[0043] The spatiotemporal alignment mapping model refers to the data fusion model constructed in this application to resolve the inconsistency in the temporal and spatial correspondence between carbon label data (business layer, discrete and batch dimensions) and equipment operating parameters (physical layer, continuous and time-series dimensions). In the time dimension, this model constructs an extended time window based on a time tolerance parameter dynamically determined by the process type. In the spatial dimension, it solves the one-to-many mapping problem between process codes and equipment codes through material flow tracking information, thereby achieving accurate matching and fusion of business data and physical data.

[0044] The time tolerance parameter, denoted as Δt, is a time margin parameter set to accommodate the offset between the carbon label timestamp and the equipment parameter timestamp. This parameter is dynamically determined based on the process type. A smaller first tolerance value is used for continuous processes, and a larger second tolerance value is used for batch processes. This is used to construct an extended time window to extract the equipment parameter sequence related to the current production batch.

[0045] Material flow tracking information refers to the real-time recording and reconstruction of the actual flow path and specific physical equipment passed by each batch of materials on the production line using digital means such as workstation records from the Manufacturing Execution System, steel passage signals from conveyor rollers, actual rack usage records, and inbound / outbound scanning records. This information is crucial for resolving ambiguities in the "one-to-many" mapping between process codes and equipment codes and for achieving precise spatial alignment of data.

[0046] The equipment operation feature vector refers to the feature vector formed after extracting statistical features from the equipment parameter sequence within a time window. The statistical features include mean, standard deviation, maximum value, minimum value, and energy consumption integral value, etc.

[0047] Fusion feature records refer to comprehensive data records formed by splicing equipment operation feature vectors with real-time carbon label data. They include both carbon label data representing carbon emission levels and feature vectors representing equipment operating status, serving as the data foundation for subsequent joint anomaly detection.

[0048] The carbon label deviation index is a quantitative indicator that characterizes the degree of deviation of the current production batch's carbon emissions from historical benchmark values, denoted as . Through formula The calculation yielded, where This represents the carbon emission value for the current batch. This serves as a historical benchmark for similar products.

[0049] Mahalanobis distance is a distance metric that considers the correlation between variables. Unlike Euclidean distance, Mahalanobis distance standardizes each dimension through a covariance matrix, thus eliminating the influence of dimensional differences and parameter correlations. In this application, it is used to quantify the deviation between the current equipment operating feature vector and the center vector of the equipment's historical normal operating range, more accurately characterizing the comprehensive abnormal state of multi-dimensional equipment parameters.

[0050] Abnormal equipment parameter indicators are quantitative indicators that characterize the degree to which the equipment's operating characteristic vector deviates from its historical normal operating range, denoted as... Mahalanobis distance is used for measurement. Mahalanobis distance can take into account the correlation between various parameters and is more suitable for anomaly detection in cases of multidimensional parameter coupling than Euclidean distance.

[0051] Dynamic threshold refers to a detection threshold that is updated periodically based on the statistical characteristics of historical data, using a formula. The calculation yielded, where This is the average of historical data. The standard deviation of historical data. This is the confidence coefficient. Unlike fixed thresholds, dynamic thresholds can adapt to gradual adjustments in production processes and baseline drift caused by equipment aging.

[0052] A carbon tag-equipment correlation anomaly refers to an abnormal state determined when both the carbon tag deviation index and the equipment parameter anomaly index simultaneously exceed their respective preset thresholds. An anomaly is only identified when both conditions are met simultaneously, ensuring a causal link between the detected anomaly and the equipment's operating condition.

[0053] A suspected anomaly refers to a state where only one condition is met: either the carbon label deviation index exceeds a preset threshold, or only the equipment parameter anomaly index exceeds a preset threshold. Suspected anomalies do not trigger automatic control processes, but are marked for further review by technical personnel to avoid overlooking potential genuine anomalies.

[0054] Anomaly contribution refers to the degree to which each device in the target device set contributes to the overall carbon tag-device association anomaly, denoted as . Through formula The calculation yielded, where The carbon emission weighting factor for the corresponding process of the equipment. This is an indicator of abnormal equipment parameters. It comprehensively considers the degree of abnormality in the equipment's parameters and the weight of the process in which the equipment operates within the overall carbon emission structure, used to accurately pinpoint the root cause of abnormal carbon emissions from among numerous devices.

[0055] The root cause device refers to the primary responsible device identified through anomaly contribution analysis that causes the carbon tag-equipment correlation anomaly. Specifically, devices whose cumulative contribution exceeds a preset contribution threshold are identified as root cause devices by sorting them in descending order of anomaly contribution and are the targets for subsequent linkage control commands.

[0056] The pre-built fault feature library refers to a fault mode-feature mapping model formed by collecting time-series data of equipment parameters from historical anomaly cases, extracting features, labeling anomaly types by process experts, and training with machine learning classification algorithms. Anomaly types include typical fault types such as decreased combustion efficiency, air-fuel ratio imbalance, bearing wear, cooling water blockage, and motor overload.

[0057] The pre-configured control strategy library refers to a set of control schemes pre-configured for various abnormal cause types. Each abnormal cause type corresponds to one or more control schemes, which are pre-configured by process experts based on production experience and process procedures.

[0058] The "ready to be issued but not yet issued" status refers to an abnormal state where steel billets have been marked as ready to be issued in the manufacturing management system (such as MES or ERP) (i.e., a production order has been issued), but due to reasons such as changes in production plans, equipment downtime due to malfunction, or data synchronization delays, the actual consumption data on the designated production equipment has not been generated or synchronized in the production execution system or process control system as expected. Batch data in this status is incomplete and needs to be removed during data processing to avoid analytical errors.

[0059] The adaptive acquisition frequency strategy refers to a strategy in which the sensor group dynamically adjusts the data acquisition frequency according to the equipment's operating status. During steady-state operation, a low-frequency acquisition mode is used. When the rate of change of a physical parameter exceeds a preset threshold, it automatically switches to a high-frequency acquisition mode. Once the rate of change recovers and continues for a preset delay, it switches back to the low-frequency acquisition mode, thus achieving a balance between data acquisition accuracy and system resource consumption.

[0060] Closed-loop verification refers to the process of continuously monitoring the carbon label deviation indicators and abnormal equipment parameter indicators of subsequent production batches after the equipment control command is executed. When the indicators of a preset number of batches return to the normal range, the abnormality is determined to be eliminated, thereby verifying the effectiveness of the abnormality handling.

[0061] The following is a brief summary of some of the innovative aspects of this application: In summary, the technical concept of this application stems from a profound understanding of the long-standing systemic technical barriers between the carbon labeling accounting system (business layer) and the equipment control system (physical layer) in the steel production process. Under the existing technical architecture, the real-time carbon label data generated by the carbon footprint accounting system (marked in the figure) The physical parameters characterizing the equipment's operating condition, collected by sensor arrays (marked in the figure as furnace temperature, gas flow rate, main motor current, etc.), belong to completely independent technical fields. The former is generated by enterprise resource planning systems or manufacturing execution systems based on production order completion times, exhibiting batch-based and discrete data characteristics. The latter is generated by data acquisition and monitoring systems or programmable logic controllers based on sensor acquisition cycles, exhibiting continuous and real-time data characteristics. When faced with carbon label anomalies, those skilled in the art typically only look for causes at the carbon accounting rule level (such as yield calculation deviations, missing outsourced processing data, inventory counting errors, etc.), without attempting to establish a technical correlation with the equipment's physical operating status. This is because the two types of data have an inherent timestamp offset in the time dimension (carbon label timestamp). Based on order completion time, device parameter timestamp Based on the sensor acquisition cycle, there is a complex many-to-many mapping between process codes and equipment codes in the spatial dimension (the same process code may correspond to multiple parallel devices, and the same device may serve multiple processes). This spatiotemporal misalignment of multi-source heterogeneous data is generally considered to be an insurmountable technical gap in this field.

[0062] The core technological contribution of this application lies in the first-ever proposal of a spatiotemporal alignment mapping model (step S3, corresponding to the spatiotemporal alignment mapping module in the figure) capable of bridging the aforementioned technological gaps. This model achieves accurate fusion of multi-source heterogeneous data through the organic combination of the following interdependent and indispensable technical means: Firstly, in the time dimension, the system obtains the start timestamp of the current production batch. and end timestamp And dynamically determine the time tolerance parameter according to the process type. (For continuous processes, a first tolerance value is used; for batch processes, a larger second tolerance value is used.) The time interval is extracted from the equipment operation parameter time series database by using the start timestamp minus the time tolerance parameter as the window start point and the end timestamp plus the time tolerance parameter as the window end point. The system employs a sequence of equipment parameters to accommodate timestamp offsets between the two systems. Secondly, in the spatial dimension, for the one-to-many mapping between process codes and equipment codes, the system uses material flow tracking information during production (including workstation records from the manufacturing execution system, transmission roller signals, and steel-passing signals from the frame) to locate the target equipment actually performing the operation within the target equipment set, thus eliminating interference from equipment belonging to the same process but not participating in the current batch. Thirdly, statistical features, including mean, standard deviation, maximum, minimum, and energy consumption integral values, are extracted from the equipment parameter sequence to generate an equipment operation feature vector. The device operation feature vector is then concatenated with the real-time carbon tag data to form a fused feature record. There is a strict technical dependency among the three techniques mentioned above: without dynamic setting of the time tolerance parameter, it is impossible to correctly extract the device parameter sequence related to the current production batch; without spatial disambiguation of material flow tracking information, irrelevant devices will be mixed into the target device set, leading to a decrease in the accuracy of subsequent anomaly tracing; without the extraction and fusion of statistical features, it is impossible to perform correlation analysis between continuous time series data and discrete carbon tag data.

[0063] Furthermore, another core technical contribution of this application lies in constructing a carbon tag-equipment joint anomaly detection mechanism (step S4, corresponding to the anomaly joint detection module in the figure) and an equipment-level anomaly tracing and location mechanism (step S5, corresponding to the anomaly tracing and location module in the figure) based on the fused feature records. These are organically combined with the generation and execution of linkage control commands (step S6, corresponding to the linkage control module in the figure) and closed-loop verification and parameter optimization (step S7, corresponding to the closed-loop verification module in the figure), forming a complete "detection-tracing-control-verification-optimization" technical closed loop. In the joint anomaly detection mechanism, the system calculates carbon tag deviation indicators characterizing the degree of carbon emission anomaly. (Formula 1) and abnormal equipment parameter indicators characterizing the degree of abnormal equipment operating conditions (Using Mahalanobis distance as the characterization), only when both the carbon label deviation index and the equipment parameter anomaly index exceed their respective preset thresholds will the condition be met. Only when (Formula 2) is the carbon label-equipment association anomaly determined. This dual-indicator joint determination logic is fundamentally different from existing methods that detect anomalies based on single-dimensional data: carbon label anomalies may be caused by various reasons such as equipment anomalies, raw material anomalies, and measurement errors; equipment parameter anomalies may also occur under normal carbon emission conditions. Only when both types of anomalies occur simultaneously does a high causal correlation confidence level exist. In the aforementioned anomaly tracing and location mechanism, the system calculates the anomaly contribution of each device in the target equipment set based on the aforementioned equipment parameter anomaly index. (Formula 3), which introduces a carbon emission weighting factor. To reflect the proportion of each process in the overall carbon emissions, the abnormality contribution is sorted in descending order, and equipment with a cumulative contribution value exceeding a preset contribution threshold is identified as the root cause equipment. The abnormal parameter patterns of the root cause equipment are then matched with a pre-set fault feature library to determine the type of abnormality. The aforementioned joint detection, source tracing, and linkage control also have strict technical dependencies: without joint judgment based on two indicators, unnecessary equipment adjustments may be triggered due to false alarms from a single dimension; without the calculation of abnormal contribution and the identification of the root cause equipment, it is impossible to determine the target of control commands; without matching the fault feature library and determining the type of abnormality, it is impossible to match a targeted control scheme from the pre-set control strategy library.

[0064] In summary, the technical solution described in this application is not a simple combination of existing carbon labeling and equipment monitoring technologies. Instead, it overcomes the technical obstacles of multi-source heterogeneous data fusion by constructing a spatiotemporal alignment mapping model, improves the accuracy of anomaly detection through a dual-indicator joint judgment mechanism, achieves precise equipment-level tracing through anomaly contribution analysis and fault feature database matching, and forms a complete closed loop from anomaly detection to equipment control through the generation and execution of linkage control commands. The aforementioned technical means are interdependent and synergistic, collectively producing outstanding technical effects that transform anomaly response from a manual processing mode to an automatic processing mode and expand carbon labeling from a post-event statistical function to a real-time feedback control function. These technical effects cannot be achieved independently by any single technical means, demonstrating the overall creative contribution of the technical solution in this application.

[0065] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0066] As stated above, the inventors of this application, through long-term and in-depth research on carbon emission management in the steel production process, have systematically analyzed the current technical situation where the carbon footprint accounting system and the equipment control system are independent of each other in the prior art, and discovered several deep-seated technical problems that restrict the refined management and control of carbon emissions.

[0067] The inventors first conducted an in-depth analysis of the fundamental reasons why carbon label data and equipment operating parameters are difficult to integrate. Through long-term practical experience, the inventors observed that the real-time carbon label data generated by the carbon footprint accounting system (such as calculated carbon emissions)... The physical parameters (such as furnace temperature, gas flow rate, and main motor current) collected by the data acquisition and monitoring system and the equipment monitoring system differ fundamentally in their data generation mechanisms: the former is generated by the enterprise resource planning system or manufacturing execution system based on the production order completion time, exhibiting batch-based and discrete data characteristics, and its timestamp reflects the order completion time rather than the actual production time; the latter is generated by the data acquisition and monitoring system or programmable logic controller based on the sensor acquisition cycle, exhibiting continuous and real-time data characteristics, and its timestamp reflects the sensor acquisition time. The inventors deeply understand that this difference in time reference is not a simple fixed offset, but rather dynamically changes with various factors such as process type, material transfer delay, and data synchronization mechanism—for continuous processes such as rolling, the material residence time on the equipment is short and relatively certain, resulting in a small time offset; for batch processes such as heating, the material residence time on the equipment is longer and fluctuates, resulting in a larger time offset. The reason existing technologies cannot effectively solve this problem is that they typically use fixed time windows or simple timestamp matching methods, which cannot adapt to the differences in time characteristics between different process types.

[0068] The inventors further analyzed the spatial alignment challenge. In actual production, there is a complex one-to-many mapping relationship between process codes and equipment codes: the same process code may correspond to multiple parallel-running devices (e.g., the roughing process in a hot rolling production line may include multiple stands), and the same device may serve multiple different production batches. Through analysis of numerous real-world cases, the inventors discovered that when a production batch experiences abnormal carbon emissions, if the associated equipment is determined solely based on the process code, all equipment under that process will be included in the analysis, including those belonging to the same process but not participating in the current batch's production. This will severely reduce the accuracy of subsequent anomaly tracing and may even lead to misjudgments. The inventors recognized that to achieve accurate spatial alignment, it is necessary to introduce tracking information that reflects the actual material flow path. However, this information is scattered across multiple data sources, such as workstation records in the manufacturing execution system, steel-passing signals from conveyor rollers, and stand usage records. Existing technologies lack an effective mechanism for integrating this information.

[0069] At the anomaly detection level, after repeated verification and theoretical derivation of existing methods for anomaly judgment based on single-dimensional data, the inventors discovered that judging solely based on carbon label deviation would misclassify carbon emission fluctuations caused by non-equipment factors such as raw material anomalies and measurement errors as anomalies requiring equipment intervention; judging solely based on equipment parameter anomalies would include minor changes in operating conditions that have not yet had a substantial impact on carbon emissions. Through in-depth consideration, the inventors creatively realized that a high degree of causal confidence exists only when carbon label anomalies and equipment parameter anomalies occur simultaneously within the same production batch. This "dual verification" judgment logic can effectively improve the accuracy of anomaly detection.

[0070] More importantly, the inventors discovered that existing technologies lack an automated closed-loop control mechanism after anomaly detection. Even if the correlation between abnormal carbon emissions and specific equipment can be identified, existing technologies cannot further determine which failure mode of the equipment caused the anomaly, and therefore cannot automatically generate targeted control commands. The inventors realized that to achieve a closed loop from anomaly detection to equipment control, it is necessary not only to identify which equipment is abnormal, but also to further determine the specific cause type of the anomaly and establish a mapping relationship between the anomaly cause type and the control scheme.

[0071] Based on the above in-depth research, the inventors of this application propose an innovative technical solution: by constructing a spatiotemporal alignment mapping model based on dynamic time tolerance parameters and material flow tracking information, the accurate fusion of carbon label data and equipment operating parameters is achieved; through a dual joint judgment mechanism of carbon label deviation indicators and equipment parameter anomaly indicators, the accuracy of anomaly detection is improved; by analyzing the anomaly contribution degree to identify the root cause equipment and matching it with a pre-set fault feature library to determine the type of anomaly cause, precise traceability at the equipment level is achieved; and through the automatic generation and execution of linkage control commands, a complete closed loop from anomaly detection to equipment control is formed.

[0072] The technical solution of this application will now be clearly and completely described with reference to the accompanying drawings. It should be noted that the following embodiments are only for explaining this application and are not intended to limit the scope of protection of this application. Those skilled in the art, based on their understanding of the technical solution of this application, can make various obvious transformations and equivalent substitutions, all of which fall within the scope of protection of this application.

[0073] The core technical idea behind the method for linking carbon label data of steel products with operating parameters of production equipment and for anomaly tracing provided in this application is to break the traditional independent technical architecture of carbon footprint accounting systems and equipment control systems. By constructing a spatiotemporal alignment mapping model, carbon label data and equipment operating parameters are deeply integrated to achieve joint detection of carbon label anomalies and equipment operating condition anomalies, equipment-level root cause tracing and location, and automatic linkage control, forming a complete technical closed loop. This method is particularly suitable for the scenario of refined carbon emission control in steel production, and can effectively solve the technical problems in existing technologies such as the disconnect between carbon label data and equipment physical state, the difficulty of spatiotemporal alignment of multi-source heterogeneous data, and the lack of a closed-loop mechanism from anomaly detection to equipment control.

[0074] like Figure 1 As shown, this embodiment includes the following specific steps: Step 100: Real-time acquisition of equipment operating parameters The first step of this method is to collect physical parameters characterizing the operating conditions of the equipment in real time by deploying sensor arrays on key equipment in steel production processes. These key equipment include, but are not limited to, heating furnaces, rolling mills, and finishing equipment, whose operating status directly affects the carbon emission levels of the products.

[0075] Specifically, the physical parameters encompass various types of process parameters, varying according to the characteristics of different equipment in different processes. For heating furnaces, the collected physical parameters include furnace temperature, gas flow rate, and flue gas oxygen content. Furnace temperature reflects the thermal state of the furnace and is a direct indicator for evaluating heating efficiency; gas flow rate is directly related to energy consumption and is positively correlated with carbon emissions; flue gas oxygen content is an important indicator for evaluating combustion efficiency—excessive oxygen content indicates excessive air and incomplete combustion, while insufficient oxygen content may lead to incomplete combustion. For rolling mills, the collected physical parameters include main motor current, rolling force, and roll speed. Main motor current reflects the mill's load level; an abnormally high current usually indicates increased equipment resistance or excessive load; rolling force and roll speed reflect the execution status of the rolling process and are closely related to product quality and energy consumption. For finishing equipment, the collected physical parameters include shear force and straightening roll pressure, which characterize the operational intensity of the finishing process. In practical applications, depending on the specific production line configuration and monitoring requirements, at least two of the above physical parameters can be selected for collection.

[0076] Furthermore, the sensor array employs an adaptive acquisition frequency strategy to balance the conflict between data acquisition accuracy and system load. The core idea of ​​this strategy is to use a low-frequency acquisition mode during steady-state operation of the device, while automatically switching to a high-frequency acquisition mode when device parameters change rapidly. This ensures data accuracy during abnormal periods while reducing the average acquisition load of the sensors and network bandwidth consumption.

[0077] More specifically, the adaptive acquisition frequency strategy includes the following sub-steps. First, during steady-state operation of the equipment, the sensor group acquires data using a first acquisition cycle, which is relatively long and is called the low-frequency acquisition mode. Second, the system monitors the rate of change of physical parameters in real time. When the rate of change exceeds a preset threshold, it indicates that the equipment may be undergoing a change in operating conditions or experiencing abnormal fluctuations. At this time, the sensor group automatically switches to a second acquisition cycle, which is relatively short and is called the high-frequency acquisition mode. Finally, when the rate of change returns to below the preset threshold and remains below it for a preset delay time, the system determines that the equipment has returned to steady-state operation, and the sensor group switches back to the low-frequency acquisition mode. It should be noted that the purpose of setting the preset delay time is to prevent frequent switching of acquisition modes near the parameter fluctuation boundary and avoid system oscillation.

[0078] The collected physical parameters are transmitted to the edge computing gateway via an industrial network. This industrial network can be an industrial Ethernet network or a fieldbus, with the specific selection determined based on site conditions. The edge computing gateway is responsible for preliminary processing and caching of the received data, including data cleaning, format conversion, and timestamp annotation, and stores the data in a time-series database of device operating parameters, providing a data foundation for subsequent spatiotemporal alignment and mapping steps.

[0079] Step 200: Extraction of Carbon Tag Association Parameters After completing the real-time acquisition of equipment operating parameters, the second step of the method in this application is to obtain the real-time carbon label data of the current production batch and its associated process list information from the carbon footprint accounting system.

[0080] The real-time carbon label data refers to the carbon footprint data identifier assigned to a batch of steel coils by a carbon footprint accounting system embedded in the production process, calculated and assigned immediately after the completion of each batch's production. This label dynamically reflects carbon emissions under specific production conditions and possesses traceability, timeliness, and uniqueness. Specifically, the real-time carbon label data includes: a product code, used to uniquely identify the batch of products; a process code, identifying each production process the product has undergone; raw material consumption and energy consumption, reflecting the material and energy inputs during production; and a carbon emission calculation value, i.e., the carbon footprint value calculated based on the aforementioned input data.

[0081] The process list information is a production process record associated with real-time carbon tag data, containing all processes performed on this batch of products and their related information. Based on the mapping relationship between the process codes in the process list information and preset equipment codes, the target equipment set corresponding to the current production batch can be determined. The target equipment set refers to the set of all equipment involved in the production of this batch of products, and is the basis for subsequent spatiotemporal alignment mapping and anomaly tracing.

[0082] Step 210: Verification of Furnaces Before Launch Optionally, before determining the target equipment set, this application also includes a pre-launch / non-launch furnace verification step to identify and eliminate invalid data. The pre-launch / non-launch furnace status is a special situation in process manufacturing industries such as steelmaking, referring to a situation where steel billets have been marked as pre-launch in the manufacturing management system (i.e., a production order has been issued), but the subsequent actual consumption data on the designated production equipment has not been generated or synchronized in the production execution system or process control system as expected. In this case, the process list information is incomplete, making effective carbon labeling and anomaly detection impossible.

[0083] Specifically, the verification steps for batches ready for shipment but not yet launched include: First, checking if there are any missing data entries for specific processes in the process list information, such as missing data for the steelmaking or continuous casting processes. Second, if missing data is found, the current production batch is determined to be in a ready-for-shipment but not yet launched state, and the system marks the batch data as invalid and terminates the execution of subsequent steps. This verification step effectively shortens the verification scope, avoids unnecessary processing of invalid data, and improves the overall efficiency of the system.

[0084] Step 300: Spatiotemporal alignment mapping steps like Figure 3 As shown, the spatiotemporal alignment mapping step is one of the core innovations of this application, aiming to construct a spatiotemporal alignment mapping model between real-time carbon tag data and physical parameters. Since carbon tag data originates from the business system, its timestamp is based on the production order completion time; while equipment operating parameters originate from the data acquisition and monitoring system or programmable logic controller, their timestamps are based on the sensor acquisition cycle. These two types of data are offset in the time dimension and have a many-to-many mapping relationship between processes and equipment in the spatial dimension, making accurate alignment difficult to achieve using traditional methods. This application achieves spatiotemporal alignment through the following sub-steps.

[0085] Step 310: Time Alignment First, obtain the start timestamp of the current production batch. and end timestamp These two timestamps, obtained from the carbon footprint accounting system or manufacturing execution system, identify the actual production period of the batch of products on the production line.

[0086] Then, determine the time tolerance parameters based on the process type. The time tolerance parameter is introduced to address the offset between the carbon tag timestamp and the equipment parameter timestamp. Since different types of processes have different production rhythms and data synchronization delays, the time tolerance parameter needs to be dynamically set according to the process type. Specifically, a first tolerance value is used for continuous processes, and a second tolerance value is used for batch processes, with the second tolerance value being greater than the first tolerance value. Continuous processes, such as rolling, involve continuous production, with short and relatively predictable material residence time on the equipment, resulting in smaller data synchronization delays; therefore, a smaller tolerance value is used. Batch processes, such as heating, have distinct batch boundaries, longer and fluctuating material residence time on the equipment, potentially leading to larger data synchronization delays; therefore, a larger tolerance value is used.

[0087] Finally, using the start timestamp minus the time tolerance parameter as the window start point and the end timestamp plus the time tolerance parameter as the window end point, the time interval is extracted from the device operation parameter time series database. The analysis includes a sequence of equipment parameters within the production batch. By extending the time window, it is ensured that all equipment parameter data related to that production batch are included in the analysis, avoiding data omissions due to time offsets.

[0088] Step 320: Spatial Alignment After time alignment, spatial alignment is also required, which maps process codes to specific equipment codes. In actual production, a process may be completed by multiple machines, and a single machine may serve multiple processes, resulting in a one-to-many mapping between process codes and equipment codes. For example, in a hot rolling production line, the roughing process may include multiple stands, the finishing process may include even more stands, and the same batch of steel may be rolled simultaneously on multiple machines. To address this, this application uses material flow tracking information during the production process to pinpoint the target equipment actually performing the operation within the target equipment set.

[0089] The material flow tracking information can come from workstation records of the manufacturing execution system, steel-passing signals from conveyor rollers, inbound / outbound scanning records, or actual rack usage records. By analyzing the actual flow path of materials during the production process, the actual equipment used in each process can be determined, thereby accurately mapping the process code to the corresponding equipment code and excluding equipment that belongs to the same process but did not participate in the current batch operation.

[0090] Step 330: Feature Fusion After completing spatiotemporal alignment, statistical feature extraction is performed on the extracted device parameter sequence to generate a device operation feature vector. This statistical feature extraction is the process of converting the original parameter sequence within a time window into a feature vector that characterizes the device's operating state.

[0091] Specifically, the process involves calculating and extracting at least three statistical features from the equipment parameter sequence within a time window: mean, standard deviation, maximum value, minimum value, and energy consumption integral value. The mean reflects the average level of the equipment parameters during that period; the standard deviation reflects the degree of fluctuation of the equipment parameters, with a larger standard deviation indicating more unstable equipment operation; the maximum and minimum values ​​reflect extreme conditions of the equipment parameters; and the energy consumption integral value is calculated by integrating the power parameters over time, reflecting the cumulative energy consumption during that period. These extracted statistical features are then combined to form an equipment operation feature vector. .

[0092] Finally, the equipment operation feature vector is concatenated with real-time carbon tag data to form a fused feature record. This fused feature record serves as the data foundation for subsequent joint anomaly detection. It contains both carbon tag data characterizing carbon emission levels and feature vectors characterizing equipment operating status, achieving deep integration of business data and physical data.

[0093] It is important to note that the device parameter sequence in step 300 consists of data collected according to the adaptive acquisition frequency strategy in step 100. This means that the data density of the device parameter sequence is low during periods of stable equipment operation, and high during periods of changing equipment operation. This data distribution characteristic enables subsequent feature extraction and anomaly detection to better capture abnormal equipment behavior.

[0094] Step 400: Anomaly Joint Detection Steps After obtaining the fused feature records, the fourth step of this application's method is to perform joint anomaly detection based on these records. The core innovation of this step lies in simultaneously considering anomalies in both the carbon tag and equipment parameter dimensions. Only when the anomaly indicators in both dimensions exceed thresholds is the carbon tag-equipment association anomaly determined. Compared to single-dimensional anomaly detection, this joint determination logic effectively reduces the false alarm rate and improves the accuracy of anomaly determination.

[0095] Step 410: Calculation of Carbon Label Deviation Index First, the carbon labeling deviation index, which characterizes the degree of carbon emission anomalies, is calculated. The carbon labeling deviation index is calculated using the following formula: (Formula 1) in, This is a carbon label deviation index, indicating the degree of deviation of the current batch's carbon emission value from the historical baseline value; The carbon emission value for the current production batch is obtained from real-time carbon label data in the fusion feature record; The historical benchmark value for similar products is usually the statistical mean or median of recent carbon emissions from similar products. This indicates the absolute value operation, which ensures that whether the carbon emission value is too high or too low, it can be detected as a deviation.

[0096] Step 420: Calculation of Abnormal Equipment Parameters Then, anomaly indicators representing the degree of abnormal equipment operating conditions are calculated. These indicators are derived from the Mahalanobis distance between the equipment's operating feature vector and the equipment's historical normal operating range. To characterize.

[0097] Mahalanobis distance is a distance metric that considers the correlation between variables. Unlike Euclidean distance, Mahalanobis distance standardizes each dimension through a covariance matrix, thus eliminating the influence of dimensions and correlations. In this application, Mahalanobis distance is used instead of Euclidean distance because there is often a correlation between equipment parameters. For example, an increase in furnace temperature is usually accompanied by an increase in gas consumption, and an increase in mill current is usually accompanied by an increase in rolling force. Mahalanobis distance can account for this correlation, avoid repeated calculations of the same anomaly, and more accurately reflect the degree of deviation in equipment status.

[0098] The covariance matrix is ​​calculated based on parameter data from the equipment's historical normal operating periods. Historical normal operating periods refer to the normal production periods excluding special periods such as maintenance and abnormalities. The further the equipment's operating characteristic vector deviates from the center of the historical normal range, the larger the Mahalanobis distance, indicating that the equipment is more likely to be in an abnormal state.

[0099] Step 430: Dynamic Threshold Generation Optionally, the preset threshold in step 400 is determined using a dynamic threshold generation method to adapt to gradual adjustments in the production process and baseline drift caused by equipment aging. The formula for calculating the dynamic threshold is: (Formula 2) in, For the preset threshold, calculate the respective thresholds for carbon label deviation indicators and abnormal equipment parameter indicators; This is the average of historical data, reflecting the central level of the indicator under normal circumstances; The standard deviation of historical data reflects the range of fluctuation of the indicator under normal conditions; is the confidence coefficient, used to control the sensitivity of the threshold.

[0100] The physical meaning of this formula is: using the mean of historical data as the center, plus a certain multiple of the standard deviation, as the boundary of the normal range. Confidence coefficient. The larger the value of the confidence coefficient, the more lenient the threshold, resulting in a lower false alarm rate but potentially a higher false negative rate; conversely, a smaller confidence coefficient indicates a lower threshold. The smaller the value, the stricter the threshold, resulting in a lower false negative rate but potentially a higher false positive rate. In practical applications, an appropriate confidence coefficient can be selected based on the specific production scenario and management needs.

[0101] Furthermore, the preset thresholds are updated on a rolling basis according to a preset cycle based on the latest historical data. For example, the thresholds can be recalculated daily or per shift based on the latest production data. This rolling update mechanism allows the thresholds to adapt to gradual adjustments in the production process, maintaining the effectiveness of the detection.

[0102] Step 440: Joint Determination After calculating the carbon label deviation index and the equipment parameter anomaly index, a joint judgment is made. The judgment logic is as follows: A carbon label-equipment correlation anomaly is determined only when the carbon label deviation index exceeds its corresponding first preset threshold and the equipment parameter anomaly index exceeds its corresponding second preset threshold. This combined "AND" logic effectively reduces false alarms caused by single-dimensional anomalies. The underlying principle is that carbon label anomalies can be caused by various factors, including equipment anomalies, raw material anomalies, and measurement errors, while equipment parameter anomalies can also occur under normal carbon emission conditions. Only when carbon label anomalies and equipment parameter anomalies occur simultaneously is there a high degree of confidence in a causal relationship between the two, thus triggering the subsequent traceability and location process.

[0103] Optionally, if only one condition is met—either the carbon label deviation index exceeds its corresponding preset threshold or the equipment parameter anomaly index exceeds its corresponding preset threshold—it is judged as a suspected anomaly and marked for review. Suspected anomalies do not trigger automatic control processes, but are recorded in the system for further analysis by technical personnel to avoid overlooking potential real anomalies.

[0104] Step 500: Abnormal Equipment Traceability and Location Steps When step 400 detects an anomaly in the carbon tag-equipment association, the fifth step of this application's method is to trace and locate the abnormal equipment, identify the root cause equipment leading to the anomaly, and determine the type of anomaly cause. This step achieves a leap from "discovering the anomaly" to "locating the root cause," providing a clear control object and control basis for subsequent linkage control.

[0105] Step 510: Calculation of Abnormal Contribution First, in response to detected carbon tag-device association anomalies, the anomaly contribution of each device in the target device set is calculated based on anomaly indicators of device parameters. The formula for calculating the anomaly contribution is: (Formula 3) in, For the first in the target device set The anomaly contribution of each device indicates the degree to which that device contributes to the overall anomaly. For the first The carbon emission weighting factor for each process corresponding to each piece of equipment reflects the proportion of that process in the overall carbon emissions. This factor is preset by process experts based on the energy consumption proportion of each process. For the first Abnormal indicators of equipment parameters for a device, namely the Mahalanobis distance value of that device; The sum of abnormal equipment parameter indices for all devices in the target device set is used as the normalization factor.

[0106] The physical meaning of this formula is: considering both the degree of equipment anomaly and its weight in carbon emissions, calculate its contribution to the overall anomaly. The introduction of a weighting factor allows equipment in processes that account for a large proportion of carbon emissions to be identified as major contributors, even if their equipment parameter anomalies are not the highest. This aligns more closely with causal relationships in actual production.

[0107] Step 520: Root Cause Device Identification Root cause devices are identified based on their abnormal contribution levels. Specifically, the target devices are sorted in descending order of their abnormal contribution levels, and then the contribution levels are accumulated from highest to lowest. Devices whose cumulative contribution exceeds a preset contribution threshold are identified as root cause devices. The preset contribution threshold is typically set to a relatively high percentage to ensure that only a small number of devices with the largest abnormal contribution are identified.

[0108] For example, if the preset contribution threshold is set to 80%, then the contribution is accumulated sequentially starting from the device with the highest contribution, stopping when the cumulative contribution reaches 80%. All devices included in the accumulation constitute the root cause device set. This method can focus on the key device with the largest abnormal contribution, avoiding indiscriminate regulation of all devices, and improving the accuracy of tracing and the targeting of regulation.

[0109] Step 530: Determine the type of cause of the anomaly After identifying the root cause device, the abnormal parameter patterns of the root cause device are matched with a pre-set fault feature library to determine the type of abnormality. The abnormal parameter patterns refer to the specific numerical combinations and fluctuation patterns presented by the device's operational feature vector.

[0110] The pre-set fault feature library is constructed as follows: First, time-series data of equipment parameters from historical anomaly cases are collected; these cases originate from past production records and fault archives. Second, features are extracted from the time-series data, and anomaly type labels labeled by process experts are obtained. Anomaly type labels can include typical fault types such as decreased combustion efficiency, combustion air-fuel ratio imbalance, bearing wear, cooling water blockage, and motor overload. Finally, a fault mode-feature mapping model is trained using machine learning classification algorithms (such as support vector machines, random forests, or neural networks) to form the pre-set fault feature library.

[0111] During online operation, the abnormal parameter patterns of the root cause devices are input into the classification model corresponding to the fault feature library, and the system outputs the matching abnormal cause type and its confidence level. The determination of the abnormal cause type provides a targeted control basis for subsequent linkage control.

[0112] Step 600: Generation and Execution of Linkage Control Commands After determining the type of anomaly, the sixth step of this application's method is to generate and execute a linkage control command. This step is a crucial link in realizing closed-loop control from anomaly detection to equipment control, and it is also a significant innovation that distinguishes this application from existing carbon labeling and accounting technologies. It enables carbon emission data to directly drive parameter adjustments in production equipment, achieving proactive optimization control of carbon emissions.

[0113] Step 610: Matching Control Plans First, based on the identified cause of the anomaly, the corresponding control scheme is matched from the pre-configured control strategy library. The pre-configured control strategy library is a collection of control schemes pre-configured for various anomaly cause types. Each anomaly cause type corresponds to one or more control schemes, which are pre-configured by process experts based on production experience and process specifications.

[0114] For example, if the cause of the abnormality is "air-fuel ratio imbalance", the corresponding control measures may include adjusting the opening of the gas valve or the frequency of the combustion fan; if the cause of the abnormality is "rolling temperature is too high", the corresponding control measures may include adjusting the cooling water flow rate; if the cause of the abnormality is "mill load is too high" or "motor overload", the corresponding control measures may include reducing the mill speed.

[0115] Step 620: Generation of Equipment Control Commands Based on the matched control scheme, equipment control instructions containing specific adjustment parameters are generated. These equipment control instructions include, but are not limited to, at least one of the following: adjustment amount of the heating furnace gas valve opening, mill speed correction value, and cooling water flow rate setpoint.

[0116] The specific adjustment parameters of the equipment control commands are calculated according to the rules in the control scheme. The control scheme usually has preset calculation logic for the adjustment direction and adjustment range, and the system calculates the specific adjustment amount based on the actual detected degree of anomaly.

[0117] Step 630: Security Verification Optionally, a safety verification step may be performed before the device control command is issued to ensure that the control command will not cause the device to exceed its safe operating range.

[0118] Specifically, the safety verification steps include: First, reading the safe operating range parameters of the root cause equipment. These parameters are derived from the equipment nameplate parameters, process cards, or control system configurations, defining the allowable upper and lower limits for each parameter. Second, verifying whether the specific adjustment parameters in the equipment control commands are within the safe operating range. Finally, if the specific adjustment parameters exceed the safe operating range, limiting the adjustment parameters to the boundary values ​​of the safe operating range; or terminating the issuance and generating a manual confirmation request for the operator to determine whether to execute.

[0119] The introduction of a safety verification step can effectively prevent equipment damage or safety accidents caused by over-regulation, thereby improving the reliability and security of the system.

[0120] Step 640: Instruction Issuance and Execution Confirmation Equipment control commands are sent to the controllers of root cause devices via an industrial control network for execution, thereby adjusting the operating status of the equipment. The controller can be a programmable logic controller (PLC) or a distributed control system. The industrial control network can employ protocols such as Industrial Ethernet, Modbus / TCP, and Profinet.

[0121] Optionally, after the device control command is issued, an execution confirmation step is also included: receiving the execution feedback signal returned by the controller of the root cause device, and confirming that the device control command has taken effect based on the execution feedback signal. The execution feedback signal may include a controller response signal, parameter readback values, execution result codes, etc. If no execution feedback signal is received within a preset time or the feedback signal indicates execution failure, the system can retry or generate an alarm to notify the operator to intervene. Through the execution confirmation step, it can be ensured that the control command is executed correctly, providing a basis for subsequent closed-loop verification.

[0122] Step 700: Closed-loop verification and parameter optimization steps The final step of this application's method is closed-loop verification and parameter optimization. This step is initiated after the equipment control command is executed. By continuously monitoring the indicator data of subsequent production batches, namely the carbon label deviation index and the abnormal equipment parameter index, the effectiveness of anomaly handling is verified, and the system parameters are optimized based on the handling results, forming a complete "detection-traceability-control-verification-optimization" technical closed loop.

[0123] Step 710: Continuous monitoring After the equipment control commands are executed, continuous monitoring is conducted on carbon label deviation indicators and abnormal equipment parameter indicators for subsequent production batches. The monitoring scope includes all subsequent production batches that have passed through the Root Cause equipment.

[0124] Step 720: Anomaly Removal Determination When the carbon label deviation index and equipment parameter anomaly index of a consecutive preset number of production batches all return to the normal range, the anomaly is determined to have been eliminated. The preset number can be set according to the production cycle and confidence requirements. Setting the requirement of consecutive batches is to avoid misjudgment caused by occasional normal batches, and to ensure that the anomaly has indeed been eliminated and is not just a random fluctuation.

[0125] Step 730: Adaptive Parameter Optimization The system parameters are updated based on the anomaly handling results to improve the system's adaptability and continuous evolution capabilities. Parameter optimization includes two aspects: First, based on the convergence of the carbon tag deviation index after anomaly handling, the preset threshold in step 400 is adjusted using an online learning method. If the anomaly handling is effective and the anomaly is quickly eliminated, the threshold can be maintained or appropriately tightened to improve detection sensitivity; if a false alarm occurs, the threshold can be appropriately relaxed (e.g., increasing the confidence coefficient k) to reduce the false alarm rate.

[0126] Second, the system receives execution feedback from the controller regarding the device control commands, and optimizes the strategy parameters of the preset control strategy library in step 600 based on the execution feedback. For example, if the adjustment range of a certain control scheme is too large, resulting in over-adjustment, the adjustment coefficient or step size of the scheme can be reduced; if the adjustment range is too small, resulting in slow anomaly elimination, the adjustment coefficient of the scheme can be increased.

[0127] Through closed-loop verification and parameter optimization steps, the method of this application can continuously learn and improve, gradually enhancing the accuracy of anomaly detection and the effectiveness of linkage control, thereby achieving continuous iterative optimization of the system.

[0128] The second embodiment of this application relates to a system for linking carbon label data of steel products with operating parameters of production equipment for monitoring and anomaly tracing, the structure of which is as follows: Figure 2 As shown, the system for monitoring and tracing anomalies in connection with carbon label data of steel products and operating parameters of production equipment includes: equipment parameter acquisition module, carbon label associated parameter extraction module, spatiotemporal alignment mapping module, anomaly joint detection module, anomaly tracing and location module, linkage control module, and closed-loop verification module.

[0129] The equipment parameter acquisition module is used to collect physical parameters characterizing the operating conditions of equipment in real time through sensor arrays deployed on key equipment in steel production processes, and transmits them to the edge computing gateway via the industrial network. This module corresponds to the function in step 100, supports an adaptive acquisition frequency strategy, and can dynamically adjust the acquisition frequency according to changes in equipment operating conditions.

[0130] The carbon label association parameter extraction module is used to obtain real-time carbon label data and associated process list information for the current production batch from the carbon footprint accounting system, and determine the target equipment set based on the mapping relationship between process codes and equipment codes. This module corresponds to the function in step 200, supports verification of furnaces that have not yet been launched, and can identify and remove invalid data.

[0131] The spatiotemporal alignment mapping module is used to construct a spatiotemporal alignment mapping model between real-time carbon tag data and physical parameters. This module extracts the equipment parameter sequence within a time window based on the start and end timestamps of the production batch and a time tolerance parameter determined by the process type. It then processes the one-to-many mapping relationship between processes and equipment based on material flow tracking information and generates fused feature records. This module corresponds to the function in step 300 and serves as a bridge connecting the business layer carbon tag data and the physical layer equipment parameters.

[0132] The anomaly joint detection module calculates carbon tag deviation indicators and equipment parameter anomaly indicators based on fused feature records, and determines a carbon tag-equipment association anomaly only when both exceed their respective preset thresholds. This module corresponds to the function in step 400, supports dynamic threshold generation and suspected anomaly marking, and can effectively reduce the false alarm rate.

[0133] The anomaly tracing and localization module is used to calculate the anomaly contribution of each device when a carbon tag-equipment association anomaly is detected to identify the root cause device, and to match the parameter anomaly pattern of the root cause device with a pre-set fault feature library to determine the type of anomaly cause. This module corresponds to the function of step 500, realizing the leap from anomaly detection to root cause localization.

[0134] The linkage control module matches a control scheme from a preset control strategy library based on the type of abnormality, generates equipment control commands, and sends these commands to the controller of the root cause device for execution. This module corresponds to the function in step 600, supporting security verification and execution confirmation to ensure the security and reliability of the control commands.

[0135] The closed-loop verification module monitors carbon label deviation indicators and abnormal equipment parameter indicators in subsequent batches to verify the effectiveness of anomaly handling, and updates preset thresholds and strategy parameters in the pre-set control strategy library based on the anomaly handling results. This module corresponds to the function in step 700, realizing the system's self-learning and continuous optimization.

[0136] The modules described above are interconnected through data interfaces to form a complete data processing and control flow. In actual deployment, the modules can run on the same computing device or be distributed across multiple computing devices, such as edge computing gateways, industrial servers, or cloud servers.

[0137] The following exemplary descriptions illustrate application scenarios of embodiments of this application. The specific application of this application is illustrated using a hot rolling production line as an example. The main equipment in a hot rolling production line includes a heating furnace and a rolling mill. Temperature sensors, gas flow meters, and flue gas analyzers are deployed in the heating furnace area, while current transformers, rolling force sensors, and speed encoders are deployed in the rolling mill area.

[0138] After the carbon footprint accounting system outputs the real-time carbon label data for a coil of steel, the system first executes step 200 to extract the furnace number and rolling mill stand number that the coil passed through, thus determining the target equipment set. Then, it executes step 300 to extract the parameter sequence of the corresponding equipment from the time-series database based on the production period of the coil, generating a fused feature record.

[0139] If step 400 detects that the carbon emission value of the coiled steel is significantly higher than normal (carbon label deviation index exceeds the standard), and at the same time the oxygen content parameter of the flue gas of a certain heating furnace is significantly abnormal (Madara distance exceeds the standard), the system determines that the carbon label-equipment association is abnormal. Step 500 identifies the heating furnace as the root cause equipment through contribution analysis, and determines the cause of the abnormality as "combustion air-fuel ratio imbalance" by matching the fault feature library.

[0140] Step 600: Based on the type of anomaly, match the control scheme of "adjusting the gas valve opening" from the control strategy library and generate an equipment control command to lower the gas valve opening. After safety verification confirms that the adjustment parameters are within the allowable range, the command is sent to the controller of the heating furnace for execution via the industrial control network. Step 700: Continuously monitor the carbon emission values ​​and equipment parameters of subsequent batches. When the indicators of several consecutive batches return to normal, the anomaly is determined to be eliminated, and the closed loop ends.

[0141] Through the above applications, this application realizes a complete closed loop from carbon tag anomaly detection to equipment-level root cause tracing and automatic linkage control, effectively solving the technical problem of the independence between traditional carbon tag accounting systems and equipment control systems, and improving the level of precision in carbon emission management of steel products.

[0142] In summary, this application, through the coordinated operation of steps 100 to 700, achieves linked monitoring and anomaly tracing of carbon label data for steel products and operating parameters of production equipment. Specifically, step 100 provides the data foundation at the physical level of the equipment, step 200 provides the carbon label data at the operational level, step 300 merges the two through spatiotemporal alignment, step 400 performs joint anomaly detection based on the merged data, step 500 performs equipment-level tracing and location of detected anomalies, step 600 converts the tracing results into equipment control actions, and step 700 verifies the control effect and optimizes system parameters. Each step forms a tight data and control flow correlation, collectively constituting a complete technical closed loop.

[0143] Compared with existing technologies, this application has the following technical advantages: By constructing a spatiotemporal alignment mapping model between carbon tag data and equipment operating parameters, it overcomes the technical bias of the two types of data being independent in existing technologies, and achieves accurate correlation between carbon tag anomalies and equipment physical status; through the generation and execution of linkage control commands, a complete closed loop from anomaly detection to equipment control is formed, reducing the anomaly response time from hours in traditional manual processing to minutes in automatic processing; through an adaptive acquisition frequency strategy, it effectively reduces sensor acquisition load and network bandwidth consumption while ensuring data accuracy; through closed-loop verification and parameter adaptive optimization, the system can continuously learn and improve, constantly enhancing the accuracy of detection and processing.

[0144] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0145] The above embodiments have the following technical effects: The steel product carbon label data and production equipment operating parameters linkage monitoring and anomaly tracing method and system provided in the above embodiments address the technical problems in the prior art, such as the independence of carbon label accounting system and equipment control system, difficulty in spatiotemporal alignment of multi-source heterogeneous data, and lack of closed-loop mechanism from anomaly detection to equipment control. Through the organic combination of multiple technical means, significant technical effects have been achieved.

[0146] Regarding real-time acquisition of equipment operating parameters, the above embodiments utilize sensor arrays deployed on key steel production equipment to collect physical parameters characterizing the equipment's operating conditions in real time. These physical parameters are then transmitted to an edge computing gateway via an industrial network, providing a physical-level data foundation for subsequent spatiotemporal alignment mapping. Furthermore, the sensor array employs an adaptive acquisition frequency strategy. During steady-state operation, it uses a low-frequency acquisition mode. When the rate of change of physical parameters exceeds a preset threshold, it automatically switches to a high-frequency acquisition mode. Once the rate of change recovers and remains for a preset delay, it switches back to the low-frequency acquisition mode. This adaptive acquisition frequency strategy enables high-density sampling data to capture abnormal transient characteristics during periods of equipment anomalies, while effectively reducing sensor load and network bandwidth consumption during steady-state periods, achieving a good balance between data acquisition accuracy and system resource consumption.

[0147] Regarding the extraction of carbon label-related parameters, the above embodiments obtain real-time carbon label data of the current production batch and its associated process list information from the carbon footprint accounting system, and determine the target equipment set based on the mapping relationship between process codes and equipment codes, thus clarifying the data range for subsequent analysis. Furthermore, before determining the target equipment set, the above embodiments introduce a pre-launch but not yet launched furnace verification step. This step identifies invalid batch data in the pre-launch but not yet launched furnace state by checking whether there are missing list data for specific processes in the process list information. The pre-launch but not yet launched furnace state refers to an abnormal state where a production order has been issued in the manufacturing management system, but no actual consumption data has been generated on the subsequent production equipment. Through this verification step, the system can effectively exclude batches with incomplete data, avoid making incorrect anomaly judgments based on incomplete data, thereby effectively shortening the verification range and improving overall analysis efficiency.

[0148] Regarding spatiotemporal alignment mapping, the above embodiments construct a spatiotemporal alignment mapping model between real-time carbon tag data and physical parameters, which is a core technical means to solve the problem of multi-source heterogeneous data fusion. In the time dimension, the system extracts the equipment parameter sequence within the time window from the equipment operation parameter time series database based on the start and end timestamps of the current production batch, combined with the time tolerance parameter determined by the process type. A smaller first tolerance value is used for continuous processes, and a larger second tolerance value is used for batch processes, thus accommodating timestamp offsets between the two systems. In the spatial dimension, for the one-to-many mapping between process codes and equipment codes, the system locks the target equipment actually performing the operation based on material flow tracking information during the production process, eliminating interference from irrelevant equipment. Regarding feature fusion, the system extracts statistical features, including mean, standard deviation, maximum, minimum, and energy consumption integral values, from the equipment parameter sequence to generate an equipment operation feature vector, and concatenates it with real-time carbon tag data to form a fused feature record. Through the above spatiotemporal alignment mapping steps, the above embodiments overcome the technical bias of existing technologies where carbon label data and equipment physical status are disconnected, and achieve deep integration of business-level carbon emission data and physical-level equipment operating condition data, laying a data foundation for subsequent joint anomaly detection.

[0149] Regarding joint anomaly detection, the above embodiments calculate carbon label deviation indicators (characterizing the degree of carbon emission anomalies) and equipment parameter anomaly indicators (characterizing the degree of equipment operating condition anomalies) based on fused feature records. A carbon label-equipment correlation anomaly is only determined when both indicators exceed their respective preset thresholds. This dual-indicator joint determination mechanism is fundamentally different from existing methods based on single-dimensional data for anomaly detection. It effectively reduces the false alarm rate caused by fluctuations in carbon label data or occasional anomalies in equipment parameters, improving the accuracy and reliability of anomaly determination. Furthermore, when only a single condition is met, it is determined as a suspected anomaly and marked for review, avoiding the omission of potential anomalies and providing a basis for subsequent manual review. In addition, the above embodiments use a dynamic threshold generation method to determine the preset threshold and update it on a rolling basis according to a preset period based on the latest historical data. This allows the detection threshold to adapt to gradual adjustments in production processes and baseline drift caused by equipment aging, exhibiting better robustness compared to fixed thresholds.

[0150] In terms of tracing and locating abnormal equipment, the above embodiments, in response to detected carbon tag-equipment-related anomalies, calculate the anomaly contribution of each device in the target equipment set based on anomaly indicators of equipment parameters. A carbon emission weighting factor is introduced to comprehensively consider the degree of anomaly and the weight percentage of the device in carbon emissions. Devices are sorted in descending order of anomaly contribution and those with a cumulative contribution value exceeding a preset contribution threshold are identified as root cause devices. This identification method focuses on the few key devices that contribute the most to the anomaly, avoiding indiscriminate control of all devices. Furthermore, the system matches the parameter anomaly patterns of the root cause devices with a pre-set fault feature library to determine the type of anomaly. This fault feature library is generated by collecting historical anomaly cases, extracting features, labeling anomaly types with process experts, and training with machine learning classification algorithms, covering common equipment anomaly patterns. Through the above anomaly tracing and locating steps, the above embodiments achieve the leap from anomaly detection to root cause location, providing clear control targets and targeted control basis for subsequent coordinated control.

[0151] Regarding the generation and execution of linkage control commands, the above embodiments match control schemes from a pre-set control strategy library based on the type of anomaly, generate equipment control commands containing specific adjustment parameters, and send the equipment control commands to the controller of the root cause device for execution via the industrial control network. The equipment control commands include adjustments to the opening of the heating furnace gas valve, mill speed correction values, and cooling water flow rate setpoints, enabling targeted control measures for different types of anomalies. Furthermore, the above embodiments perform a safety verification step before issuing the command to check whether the specific adjustment parameters in the equipment control command are within the safe operating range of the root cause device. If they exceed this range, a limiting process is performed or a manual confirmation request is generated, effectively preventing equipment damage or safety accidents caused by excessive control. In addition, the above embodiments perform a confirmation step after issuing the command, receiving an execution feedback signal from the controller to confirm that the equipment control command has taken effect, ensuring reliable execution of the control command. Through the above-mentioned linkage control steps, the above embodiments form a complete closed loop from anomaly detection to equipment control, expanding carbon tag accounting from a post-event statistical function to an active management function that can provide real-time feedback and control of the production process, and significantly shortening the anomaly response time from hours in the traditional manual processing mode to minutes in the automatic processing mode.

[0152] Regarding closed-loop verification and parameter optimization, the above embodiments continuously monitor the carbon label deviation indicators and abnormal equipment parameter indicators of subsequent production batches after the equipment control commands are executed. When the indicators of a preset number of consecutive batches recover to the normal range, the anomaly is determined to be eliminated. This determination method based on the recovery of indicators from multiple consecutive batches is more robust than that based on a single batch, avoiding misjudgments caused by accidental fluctuations. Furthermore, the system adjusts preset thresholds using an online learning method based on the anomaly handling results and optimizes the strategy parameters of the preset control strategy library based on the controller's execution feedback. This allows the system to continuously learn and improve, continuously enhancing the accuracy and efficiency of carbon label anomaly detection and processing, and achieving adaptive evolution of the system.

[0153] In summary, the above embodiments, through the organic combination and synergistic cooperation of technologies such as real-time acquisition of equipment operating parameters, extraction of carbon tag-related parameters, spatiotemporal alignment mapping, joint anomaly detection, tracing and locating abnormal equipment, generation and execution of linkage control commands, and closed-loop verification and parameter optimization, have constructed a complete technical system for the linkage monitoring and anomaly tracing of carbon tag data and production equipment operating parameters in the steel industry. This technical system overcomes the limitations of the existing technology's independent technical architecture between the carbon tag accounting system and the equipment control system, achieving precise correlation between carbon tag anomalies and equipment physical status, precise equipment-level tracing and location, and an automated closed loop from anomaly detection to equipment control. This provides effective technical support for the refined management of carbon emissions in the steel production process.

[0154] It should be noted that those skilled in the art should understand that the implementation functions of each module shown in the above-described implementation of the system for monitoring and tracing anomalies in the linkage between carbon label data of steel products and operating parameters of production equipment can be understood with reference to the relevant description of the aforementioned method for monitoring and tracing anomalies in the linkage between carbon label data of steel products and operating parameters of production equipment. The functions of each module shown in the above-described implementation of the system for monitoring and tracing anomalies in the linkage between carbon label data of steel products and operating parameters of production equipment can be implemented by a program (executable instructions) running on a processor, or by specific logic circuits. If the above-described system for monitoring and tracing anomalies in the linkage between carbon label data of steel products and operating parameters of production equipment is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of this application, in essence, or the part that contributes to the prior art, 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 methods described in the various embodiments of this application. The aforementioned storage media include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks. Therefore, the embodiments of this application are not limited to any specific hardware and software combination.

[0155] Accordingly, this application also provides a computer storage medium storing computer-executable instructions, which, when executed by a processor, implement the various method implementations of this application.

[0156] Furthermore, this application also provides a system for linking carbon label data of steel products with operating parameters of production equipment for monitoring and anomaly tracing. This system includes a memory for storing computer-executable instructions and a processor. The processor is used to implement the steps in the above-described method embodiments when executing the computer-executable instructions stored in the memory. The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The aforementioned memory can be read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or solid-state drive, etc. The steps of the methods disclosed in the embodiments of this application can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0157] It should be noted that in this patent application, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. In this patent application, if it refers to performing an action according to an element, it means performing the action at least according to that element, including two cases: performing the action only according to that element, and performing the action according to that element and other elements. Expressions such as "multiple," "repeatedly," and "various" include two, two times, two kinds, and more than two, more than two times, and more than two kinds.

[0158] All documents mentioned in this application are considered to be incorporated in their entirety into the disclosure of this application so that they can serve as a basis for modifications if necessary. Furthermore, it should be understood that after reading the foregoing disclosure of this application, those skilled in the art can make various alterations or modifications to this application, and these equivalent forms also fall within the scope of protection claimed in this application.

Claims

1. A method for monitoring and tracing abnormality of linkage between carbon label data of steel products and operating parameters of production equipment, characterized in that, Includes the following steps: S1: Through sensor arrays deployed on key equipment in steel production processes, physical parameters characterizing the operating conditions of the equipment are collected in real time and transmitted to the edge computing gateway via the industrial network; S2: Obtain real-time carbon label data of the current production batch and its associated process list information from the carbon footprint accounting system, and determine the target equipment set based on the mapping relationship between process codes and equipment codes; S3: Construct a spatiotemporal alignment mapping model between the real-time carbon tag data and the physical parameters; extract the equipment parameter sequence of each device in the target equipment set within the corresponding time window from the equipment operation parameter time series database; extract statistical features from the equipment parameter sequence to generate an equipment operation feature vector; and concatenate it with the real-time carbon tag data to form a fusion feature record. S4: Based on the fusion feature record, calculate the carbon tag deviation index and the equipment parameter anomaly index. Only when both exceed their respective preset thresholds is it determined that the carbon tag-equipment association is abnormal. S5: In response to the carbon tag-equipment association anomaly, calculate the anomaly contribution of each device to identify the root cause device, and match the parameter anomaly pattern of the root cause device with a preset fault feature library to determine the anomaly cause type; S6: Match a control scheme from the preset control strategy library according to the type of abnormal cause, generate equipment control instructions and send them to the controller of the root cause device for execution; S7: Monitor the carbon label deviation index and the abnormal equipment parameter index of subsequent production batches. When the index of a preset number of consecutive batches returns to normal, the abnormality is determined to be eliminated, and the preset threshold is updated according to the abnormality handling result.

2. The method of claim 1, wherein, The construction of the spatiotemporal alignment mapping model in S3 also includes: The corresponding time window is determined based on the start and end timestamps of the current production batch, combined with the time tolerance parameter determined by the process type. Given the one-to-many mapping between process codes and equipment codes in the process list information, the target equipment that actually performs the operation is located in the target equipment set based on the material flow tracking information in the production process.

3. The method of claim 1, wherein, In step S1, the sensor group employs an adaptive acquisition frequency strategy, specifically including: Low-frequency data acquisition mode is used during steady-state operation of the equipment; The rate of change of the physical parameters is monitored in real time, and when the rate of change exceeds a preset threshold, the system automatically switches to high-frequency acquisition mode. When the rate of change recovers to below the preset change threshold and remains below it for a preset delay time, the system switches back to the low-frequency acquisition mode. The device parameter sequence in S3 consists of data collected according to the adaptive acquisition frequency strategy.

4. The method of claim 1, wherein, In S1, the physical parameters include at least two of the following: furnace temperature, gas flow rate and oxygen content of flue gas in the heating furnace, main motor current, rolling force and roll speed of the rolling mill, and shearing force and straightening roll pressure of the finishing equipment.

5. The method of claim 1, wherein, In step S2, before determining the target equipment set, a step of verifying quasi-launching but not yet launched furnaces is also included: Check whether there are any missing list data for a specific process in the process list information; If any data is missing, the current production batch is determined to be in a state of being ready for launch but not yet launched, marked as invalid data, and subsequent steps are terminated. The "ready-to-launch but not yet launched" status refers to a state in which a production order has been issued in the manufacturing management system, but no actual consumption data has been generated on the subsequent production equipment.

6. The method of claim 2, wherein, In step S3, the device parameter sequence of each device in the target device set within the corresponding time window is extracted from the device operation parameter time series database, specifically including: Obtain the start timestamp of the current production batch. and end timestamp ; determining the time tolerance parameter according to the process type wherein a first tolerance value is used for continuous processes and a second tolerance value is used for batch processes, and the second tolerance value is greater than the first tolerance value. Using the start timestamp minus the time tolerance parameter as the starting point of the window, and the end timestamp plus the time tolerance parameter as the ending point of the window, a time interval is extracted from the device operating parameter time series database. The sequence of device parameters within.

7. The method of claim 1, wherein, In step S3, extracting statistical features from the device parameter sequence to generate a device operation feature vector includes: The device parameter sequence within the time window is calculated, and at least three statistical features are extracted, including the mean, standard deviation, maximum value, minimum value, and energy consumption integral value. The extracted statistical features are combined to form the device operation feature vector.

8. The method of claim 1, wherein, In step S4, the carbon label deviation index is calculated using the following formula: wherein, is the carbon label bias indicator, is the carbon emission value of the current production batch, is the historical benchmark value of the same product, denotes the absolute value operation; The device parameter anomaly index is represented by a Mahalanobis distance between the device operation feature vector and the historical normal operation interval of the device .

9. The method of claim 8, wherein, The preset threshold in S4 is determined using a dynamic threshold generation method, and the formula is: in, The preset threshold, The average of historical data. The standard deviation of historical data, Confidence coefficient; The preset threshold is updated on a rolling basis according to a preset period based on the latest historical data.

10. A steel product carbon label data and production equipment operation parameter linkage monitoring and abnormality tracing system, characterized in that, include: The equipment parameter acquisition module is used to collect physical parameters that characterize the operating conditions of the equipment in real time through a sensor group deployed on key equipment in steel production processes, and transmit them to the edge computing gateway through the industrial network. The carbon label association parameter extraction module is used to obtain real-time carbon label data of the current production batch and its associated process list information from the carbon footprint accounting system, and determine the target equipment set based on the mapping relationship between process code and equipment code; The spatiotemporal alignment mapping module is used to construct a spatiotemporal alignment mapping model between the real-time carbon tag data and the physical parameters. It extracts the equipment parameter sequence within the time window based on the start and end timestamps of the production batch and the time tolerance parameter determined by the process type. It processes the one-to-many mapping relationship between the process and the equipment based on the material flow tracking information and generates a fused feature record. The anomaly joint detection module is used to calculate the carbon tag deviation index and the equipment parameter anomaly index based on the fused feature record, and to determine the carbon tag-equipment association anomaly only when both exceed their respective preset thresholds. An anomaly tracing and localization module is used to calculate the anomaly contribution of each device to identify the root cause device when the carbon tag-equipment association anomaly is detected, and to match the parameter anomaly pattern of the root cause device with a preset fault feature library to determine the anomaly cause type. The linkage control module is used to match a control scheme from a preset control strategy library according to the type of abnormal cause, generate equipment control instructions, and send the equipment control instructions to the controller of the root cause device for execution; The closed-loop verification module is used to monitor the carbon label deviation index and equipment parameter anomaly index of subsequent batches to verify the anomaly handling effect, and update the preset threshold and the strategy parameters of the preset control strategy library according to the anomaly handling results.

Images