Industrial safety alarm event recording and tracing method based on graphic control system

Abstract

The invention provides an industrial safety alarm event recording and tracing method based on a graphic control system, and relates to the technical field of industrial automation. The method comprises the following steps: carrying out undersampling processing on a collected original alarm signal, and generating a reconstructed alarm signal by using a sparse reconstruction algorithm; dividing the reconstructed alarm signal into alarm events of different alarm levels and alarm areas according to a pre-configured area division rule and an alarm type classification rule; meanwhile, the grade, the area, the occurrence time and the processing state of the alarm event are associated with PLC data points and written into a historical database; and after a query instruction of a user is received, the historical record of the alarm event is extracted from the database, and visual tracing is carried out on a graphic control system interface. According to the method, while the pressure of mass alarm data transmission and storage is reduced, key features of alarm signals can be accurately reserved, automatic recording and rapid tracing of multi-source heterogeneous industrial alarms are achieved, and the safety alarm management efficiency and reliability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial automation, and in particular to an industrial safety alarm event recording and tracing method based on a graphic control system. Background Art

[0002] In large-scale industrial production processes, factories usually deploy a large number of sensors and PLCs (programmable logic controllers) to obtain and transmit production operation information in real time, and implement industrial alarm monitoring within the management platform (such as the graphic control system). However, with the increasing complexity of industrial production processes, the number of alarm signals has increased sharply, and the sampling cycles and data formats of different plant areas or different work sections are not the same, resulting in traditional alarm management solutions facing problems such as high transmission bandwidth pressure, difficulty in quickly storing and analyzing massive data, and difficulty in effectively integrating multi-source information. In addition, if only high-frequency sampling under a single working condition is used to capture alarm signals, a large amount of redundant data will often be generated, and the alarm correlation across regions or plant areas cannot be fully utilized. False alarms or missed alarms are prone to occur, and there is a lack of tools and means for fine tracing and visual tracking of historical records, which brings great challenges to industrial safety operation and maintenance. Summary of the invention

[0003] In view of the shortcomings of the prior art, the present invention provides an industrial safety alarm event recording and tracing method based on a graphic control system, including: Collecting an original alarm signal, and performing undersampling processing on the original alarm signal to generate a compressed alarm signal; reconstructing the compressed alarm signal by using a sparse reconstruction algorithm to generate a reconstructed alarm signal; According to preset area division rules and alarm type classification rules, the reconstructed alarm signal is divided into alarm events of different alarm levels and different alarm areas; According to the divided alarm events, the alarm level, area information, occurrence time and processing status of the alarm event are automatically recorded in the historical database through the pre-associated PLC data points to form the alarm event history record; Receive a user's query instruction, extract corresponding alarm event history records from the history database according to the query instruction, and perform visual tracing display on the graphics control system interface.

[0004] As an optional implementation, it also includes: When configuring the graphics control system, a domain partition library is created, and a domain identifier is assigned to each heterogeneous data source in the domain partition library, and a mapping relationship is established between the domain identifier and the corresponding sensor list, sampling frequency parameter, and production unit or plant area identifier; In response to a reconstructed alarm signal within a domain exceeding a preset threshold, the reconstructed alarm signal and the historical reference data corresponding to the domain are sent to a cross-domain fusion classification module to obtain a preliminary alarm classification result for the domain. After obtaining the classification result, the domain identifier and the corresponding preliminary alarm classification result are written into a historical database.

[0005] As an optional implementation, it also includes: A distribution difference measurement unit is set in the cross-domain fusion classification module, and the distribution difference measurement unit calculates the distribution difference degree based on the characteristics of the reconstructed alarm signals collected from each source domain and the target domain in the domain division library and based on a preset divergence calculation method; The cross-domain fusion classification module is also provided with a discount coefficient calculation unit for generating reliability or discount coefficient according to the distribution difference; After completing the calculation of distribution difference and discount coefficient, the alarm judgment results of each source domain are weighted and fused, and the comprehensive alarm category and alarm level are output. A comprehensive judgment result is obtained through a preset synthesis formula. When the judgment result is higher than the preset fusion threshold, the comprehensive alarm category and alarm level are written into the historical database.

[0006] As an optional implementation, it also includes: A trust classification model is configured in the cross-domain fusion classification module, wherein the trust classification model includes a neighbor search unit and a composite class determination unit, wherein the neighbor search unit is used to identify the concentration of fault labels based on the neighborhood distribution of the reconstructed alarm signal in the high-dimensional feature space; The comprehensive alarm category obtained in response to the preset synthesis formula includes multiple fault labels, which triggers the composite category determination unit to identify the alarm event as a composite alarm, and records the source domain information, alarm category, alarm area and timestamp of the composite alarm to the historical database, so that the graphics control system interface can trace the source and display the composite alarm event information when receiving a user query instruction.

[0007] As an optional implementation, it also includes: Setting a PLC attribute domain evidence source in the cross-domain fusion classification module and storing a health record table for each PLC in a historical database; The health record table includes the scan cycle average, communication timeout count and hardware self-test result, which are used to indicate the operating status of the PLC; Before calculating the distribution difference of the reconstructed alarm signal and fusing it with the synthesis formula, querying the health record table of the PLC attribute domain evidence source and generating the corresponding attribute domain confidence; In response to the health record indicating that the PLC is in a normal operating state, increasing support for the alarm category; In response to the health record table indicating that the PLC communication timeout or hardware failure is detected, the support or confidence of the alarm category is reduced, and the fusion result corrected by the PLC attribute domain evidence source is written into the historical database.

[0008] As an optional implementation, it also includes: In the sparse reconstruction algorithm, the target PLC is set as the PLC that generates a compressed alarm signal and triggers an alarm state through the graphic control system; Setting an association relationship table for the signal association between each PLC in the historical database to store association parameters obtained based on similarity or mutual information; A multi-source dictionary management module is configured in the sparse reconstruction algorithm to load the historical waveform basis functions of the target PLC and the associated PLC; In response to the health value of the target PLC being less than a preset threshold, the multi-source dictionary management module references a corresponding pattern from the waveform basis function of the associated PLC and reduces the weight of the target PLC local basis function in the reconstruction calculation to generate the reconstruction alarm signal; In response to the health value of the target PLC being greater than or equal to the preset threshold, the multi-source dictionary management module performs reconstruction calculation using the local basis function of the target PLC and outputs the reconstruction alarm signal; After the reconstruction alarm signal is generated, the dictionary call information of the multi-source dictionary management module is written into the historical database together with the health records of the target PLC and the associated PLC.

[0009] As an optional implementation, it also includes: Install a vibration trigger for the monitoring object where the target PLC is located, and establish a vibration reference table for each vibration mode in the historical database to store the preset excitation frequency and theoretical response curve; When the graphics control system is in a non-critical operation period, an active vibration instruction is sent to the vibration trigger, and sampling data of the target PLC in the vibration period is obtained; After comparing the sampled data with the theoretical response curve, if the deviation between the two is greater than a threshold, an active vibration detection abnormality is marked in the health record table corresponding to the target PLC.

[0010] As an optional implementation, it also includes: A vibration basis function library is provided in the sparse reconstruction algorithm for storing standard waveform basis functions corresponding to each excitation mode in the vibration reference table; In response to an active vibration detection abnormality flag of a target PLC being written, applying a discount coefficient to a local basis function used by the target PLC in a reconstruction calculation, and increasing the priority of the associated PLC basis function in a multi-source dictionary management module; After the reconstruction is completed, the reference information of the vibration base function library and the active vibration detection abnormality mark of the target PLC are written into the historical database.

[0011] As an optional implementation, it also includes: Setting a vibration excitation cycle configuration item in the database or configuration file to periodically trigger the vibration trigger and collect vibration response data of the target PLC; In response to detecting that the vibration response of the target PLC deviates from the theoretical response curve by an amplitude or frequency exceeding a preset range for multiple times during the excitation cycle, automatically reducing the health value of the target PLC to below a predetermined threshold; After the health value decreases, when cross-domain reconstruction is performed on the compressed alarm signal uploaded by the target PLC based on the sparse reconstruction algorithm, other PLC basis functions whose health values ​​are greater than or equal to a predetermined threshold are called to compensate for the data distortion of the target PLC, and the final reconstruction result is written into the historical database.

[0012] Compared with the prior art, this application uses undersampling and sparse reconstruction to significantly reduce the volume of alarm data while ensuring the fidelity of the alarm signal. At the same time, it combines the preset regional division rules and alarm type classification rules to quickly and accurately map the reconstructed alarm signal to different alarm areas and alarm levels. More importantly, while recording the alarm event level, regional information, timestamp and processing status, the present invention forms a complete alarm history database through the association relationship with the PLC data points, and supports visual backtracking for different query requirements. It not only improves the processing efficiency of multi-source heterogeneous alarm signals in complex industrial sites, but also provides a reliable data foundation for subsequent industrial safety analysis and equipment maintenance optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A flow chart of an industrial safety alarm event recording and tracing method based on a graphic control system provided in an embodiment of the present application; Figure 2 A schematic diagram of a naming standard for various detectors provided in an embodiment of the present application; Figure 3 A schematic diagram of an example of naming and annotating detector-related points provided in an embodiment of the present application; Figure 4 A schematic diagram of a health record maintenance process using a vibration unit provided in an embodiment of the present application. DETAILED DESCRIPTION

[0014] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0015] See also Figure 1 , Figure 1 The flowchart of the industrial safety alarm event recording and tracing method based on the graphic control system provided in the embodiment of the present application includes steps S101 to S104, wherein: S101: collecting an original alarm signal, and performing undersampling processing on the original alarm signal to generate a compressed alarm signal; reconstructing the compressed alarm signal by using a sparse reconstruction algorithm to generate a reconstructed alarm signal; S102: dividing the reconstructed alarm signal into alarm events of different alarm levels and different alarm areas according to preset area division rules and alarm type classification rules; S103: According to the divided alarm events, the alarm level, area information, occurrence time and processing status of the alarm event are automatically recorded in the historical database through the pre-associated PLC data points to form an alarm event history record; S104: receiving a query instruction from a user, extracting corresponding alarm event history records from the history database according to the query instruction, and performing visual tracing display on the interface of the graphic control system.

[0016] When deploying the safety alarm event recording and tracing method based on the graphic control system described in the present invention at an industrial site, it is usually necessary to comprehensively consider multiple factors such as hardware environment, network communication, data storage, algorithm implementation, and human-computer interaction. The following takes the implementation of over-temperature alarm monitoring in a petrochemical production device as an example to illustrate the specific engineering implementation steps of the present invention.

[0017] First, at the hardware level, key equipment in petrochemical plants (such as reactors, separation towers, pipeline valves, etc.) are usually equipped with several temperature, pressure, flow and other sensors, and the data collected by the sensors are preliminarily processed or the switch quantity is judged through industrial-grade PLCs (such as Siemens S7 series, Schneider Modicon series or Rockwell Allen-Bradley, etc.). Real-time data exchange can be carried out between PLC and the host system through industrial Ethernet protocols (such as PROFINET, EtherNet / IP, Modbus TCP or OPC UA, etc.). In order to reduce network bandwidth pressure, reduce storage volume, and ensure timely capture of alarm information under critical working conditions, the present invention deploys undersampling and sparse reconstruction modules at the data acquisition end or in the middle edge computing node.

[0018] Specifically, a dedicated signal processing program is run on the field data acquisition station (an industrial computer or edge gateway device can be used), which regularly obtains the original alarm signal through the communication interface of the PLC. When the system detects that the temperature value of a certain measuring point exceeds the alarm threshold, it is considered to be in an "alarm state" and performs undersampling according to the sampling period set according to process needs. For example, for the overheating reaction stage with drastic temperature changes, the sampling period can be dynamically adjusted to 50ms; for the stable operation stage, it can be relaxed to 100ms or longer to reduce the amount of data without affecting the capture of key features. The undersampling process is usually combined with digital filtering and denoising to avoid unnecessary negative impacts of interference peaks on the reconstruction algorithm.

[0019] After obtaining the undersampled data, the signal processing program will call the sparse reconstruction algorithm library to complete the reconstruction of the alarm signal. The algorithm library can be implemented based on C / C++, Python or other high-performance languages ​​to ensure real-time performance in industrial field environments. The algorithm can use orthogonal matching pursuit (OMP), K-SVD based on dictionary learning, and other reconstruction methods based on compressed sensing theory.

[0020] In order to improve the reconstruction effect, it is necessary to establish a dictionary library that conforms to the characteristics of the production process during system configuration. The dictionary library contains common mode basis functions of temperature changes (such as slow rise, rapid jump, oscillation, etc.), and the dictionary library can also be continuously optimized through online learning during the initial operation stage. After being processed by the above sparse reconstruction algorithm, the reconstructed alarm signal is closer to the original signal in terms of time resolution and amplitude accuracy, thus providing more comprehensive data support for subsequent alarm classification and trend analysis.

[0021] After the alarm signal is reconstructed, the system will determine the alarm event information according to the pre-configured regional division and alarm type classification rules. Taking the petrochemical plant as an example, the entire plant may be divided into raw material storage area, reaction area, heat exchange area and finished product tank area according to geographical location or process flow; at the same time, different level threshold values ​​from level I to level III can be set in the temperature alarm dimension to distinguish between warning, failure and emergency. When the reconstructed alarm signal corresponding to a sensor exceeds the level II threshold, the system will automatically generate a level II alarm event and identify the corresponding area. When implementing this process, it is usually necessary to maintain a mapping relationship table between "measurement point-area-alarm threshold-alarm level" in the system database or configuration file; once the alarm exceeds the limit, the system can query the relationship table to quickly determine the area and level of the alarm. In order to ensure the flexibility of project implementation, operators can also dynamically modify the mapping relationship on the background interface of the graphic control system to adapt to the adjustment needs of the alarm threshold in different production stages or seasonal working conditions.

[0022] After completing the alarm event determination, the system will automatically write the core data of the alarm event (including alarm area, alarm level, generation time, alarm type, and the PLC data point ID associated with it) into the historical database. The historical database here can be a traditional relational database (such as Oracle, SQL Server, MySQL), or a time series database (such as InfluxDB, TimescaleDB) to better meet the needs of high-frequency writing and time series analysis.

[0023] In specific engineering implementation, the high availability and scalability of the database also need to be considered. For example, for large petrochemical complexes, it is recommended to deploy a dual-machine or cluster database solution to prevent production data loss caused by single point failures. At the same time, when the system writes alarm records, it can also record user interaction information related to alarm processing, such as which operator confirmed or processed the alarm and when, as well as the operating instructions during the processing process, the interlocking measures used, etc.

[0024] In the upper interface of the graphic control system (such as SCADA or DCS system), the visual tracing function of the present invention is mainly realized by retrieving and presenting the alarm events stored in the historical database. For the petrochemical site, the P&ID (process flow chart) or 3D digital twin model of the plant can be associated with the alarm data. When the user clicks on an area or equipment, the system will display a list of the most recent alarm records of the equipment on the interface. The list includes the alarm time, alarm level, processing status, and whether linkage measures have been taken. If the user needs a more detailed traceability analysis, click on the alarm record to enter the trend viewing mode, and retrieve the temperature curve (or other related parameters such as pressure, flow, etc.) of the alarm several minutes to hours before and after the trigger, so as to evaluate the causes and consequences of the failure, and finally provide decision makers with reliable clues for root cause analysis of the failure.

[0025] In addition, in some application scenarios, operators may also need to perform statistical analysis and report output on alarm events, such as counting the total number of alarms in each area within a certain period of time, the proportion of alarms of each level, and the average processing time and response time. When the present invention is implemented in an engineering project, a statistical function module can be added at the database level or the graphics control system level, and a large number of historical alarms can be quickly summarized through aggregate query or big data analysis technology, and then a visual report can be generated in the form of a bar chart, pie chart or line trend chart, or email or SMS reports can be sent to the management on a regular basis. These functions can help factory managers understand the health of equipment and the operating status of the process, and provide a scientific basis for subsequent production improvements, equipment pre-maintenance and personnel scheduling.

[0026] In this way, the present application effectively realizes the efficient compression, accurate reconstruction, automatic event division, historical record management and visual tracing of industrial site safety alarm signals. Among them, undersampling and sparse reconstruction technology not only reduces the consumption of transmission and storage resources, but also ensures the preservation and restoration of key features of alarm waveforms through appropriate dictionary design and online learning mechanisms. The function of automatically recording alarm events into a historical database and visually querying them improves the response efficiency and traceability analysis capabilities of operation and maintenance personnel to alarm events.

[0027] Exemplarily, the alarm group division rule may be: Secondary group division rule: There is only one GMS secondary group division rule: There are as many secondary alarm groups as there are types of equipment in the menu, such as the detector corresponds to the alarm group GMS_DT, the alarm light corresponds to the alarm group GMS_LAU, etc.; Three-level group division rules: For each second-level group, there will be several related building alarm groups. For example, for the detector alarm group GMS_DT, there will be the following three corresponding third-level alarm groups: GMS_FAB1_DT (all FAB1 detectors will belong to this group), GMS_FAB2_DT (all FAB2 detectors will belong to this group), GMS_SIH4YD_DT (all silane station detectors will belong to this group); Four-level group division rules: For each three-level group, if it needs to be divided into smaller areas, it can be divided again. For example, the three-level group GMS_FAB1_DT represents all detectors of FAB1, and FAB1 has two floors, so it can be divided into GMS_FAB1_F1_DT (all detectors on the first floor of FAB1 will belong to this group) and GMS_FAB1_F2_DT (all detectors on the second floor of FAB1 will belong to this group); Division rules of five-level groups: For each four-level group, if it needs to be divided into smaller areas, it can be divided again. For example, the four-level group GMS_FAB1_F1_DT represents all detectors on the first floor of FAB1. There are many machines on the first floor of FAB1, and there are many detectors in (or near) each machine. So it can be divided again according to needs, such as GMS_FAB1_F1_CVD1_DT (all CVD1-related detectors on the first floor of FAB1 will belong to this group).

[0028] Exemplary, the naming rules for alarm groups related to detectors are: GMS_DT (second-level alarm group naming), GMS_large area_DT (third-level alarm group naming, the large area is generally the building name), GMS_large area_medium area_DT (fourth-level alarm group naming, the medium area is generally the floor number), GMS_large area_medium area_small area_DT (fifth-level alarm group naming, the small area is generally a small room), GMS_large area_medium area_small area_smaller area_DT (sixth-level alarm group naming).

[0029] Except for the second-level alarm group GMS_DT, the subsequent specific alarm groups can be divided layer by layer according to the area. The number of areas in different factories is different, which means that the depth of the specific alarm groups is also different.

[0030] Among them, GMS: General / Gas Monitoring System, general / gas monitoring system; FAB: Fabrication Facility, manufacturing workshop or factory; DT:Detector,detector; LAU: Local Alarm Unit, local alarm unit (such as alarm light or siren); CVD: Chemical Vapor Deposition, chemical vapor deposition equipment; SIH4YD:Silane Yard, silane station or silane storage site; F1, F2: Floor 1, Floor 2, indicating the first floor and the second floor.

[0031] For example, see Figure 2 , Figure 2 A schematic diagram of a naming convention for various detectors provided in an embodiment of the present application.

[0032] For example, see Figure 3 , Figure 3 A schematic diagram of an example of naming and annotating detector related points provided in an embodiment of the present application.

[0033] As an optional implementation, although the industrial safety alarm event recording and tracing method based on the graphic control system disclosed in the present invention can reduce the burden of data transmission and storage by compressing and sparsely reconstructing the original alarm signal, it can retain the main characteristic information of the alarm signal. However, in some complex industrial occasions, especially when the production process involves multiple types of devices, cross-regional or cross-plant linkage, the alarm signal often comes from multiple heterogeneous data sources, and these data sources may be distributed in different production units or different plant areas (ie, "multi-source domains"). Due to differences in process flow, operating methods or environmental conditions, the distribution of data from different sources is inconsistent, which brings certain challenges to the accurate determination and classification of alarm events.

[0034] In order to further improve the accuracy of alarm event recognition and classification in multi-source heterogeneous data scenarios, the present invention can perform cross-domain auxiliary classification and fusion decision-making based on the alarm signals obtained by sparse reconstruction.

[0035] In specific implementation, in a typical industrial production environment, different factory areas or production lines often deploy their own sensors and PLC systems. The alarm data output by these systems may differ in signal sampling frequency, data format and measurement point definition, and are typical heterogeneous data sources.

[0036] To facilitate subsequent cross-domain fusion processing, the present invention first divides these different data sources into "domains" during system configuration. Each domain can correspond to a factory area, a production line or a subsystem (such as a heating furnace domain, a reaction area, a storage tank domain, etc.), and completes preliminary undersampling and sparse reconstruction operations locally or at the edge.

[0037] When the system detects a potential alarm situation in a domain (such as temperature, pressure, flow rate and other parameters exceeding the limit), the reconstructed alarm signal of the domain will be input into the cross-domain fusion classification module together with the historical reference data (annotated fault samples, normal samples, etc.). The purpose of this is to improve the judgment ability of the current alarm signal with the help of the fault knowledge and classification models accumulated in other domains (Source Domain) when the current target domain (Target Domain) lacks sufficient training data or annotation information.

[0038] Furthermore, in order to better utilize data from multiple source domains to assist in the classification of target domain alarms, the present invention takes the following steps: compare the feature distribution of each source domain (such as the time domain features, frequency domain features, statistical features, etc. of the alarm signal) with the target domain; use KL divergence (KL-Divergence), MMD (Maximum Mean Discrepancy) or JS divergence to measure the distribution difference; set "reliability" or "discount coefficient" according to the size of the difference. If the distribution similarity between a source domain and the target domain is higher, its reliability is higher, otherwise the discount coefficient is larger. In this way, in the scenario where multiple source domain data are jointly assisted, the contribution of each source domain can be dynamically weighted to reduce the error caused by the large distribution difference and improve the classification accuracy of cross-domain migration.

[0039] After obtaining the judgment result of each source domain on the "current target alarm signal", the present invention integrates the results of each source domain through evidence theory fusion to obtain a more robust alarm classification output.

[0040] Specifically: Each source domain classifies the target domain signal based on its existing fault / alarm model, such as determining whether it is "temperature warning", "temperature fault", "valve switch abnormality", "comprehensive fault", etc. For each source domain, the confidence (or “basic probability distribution”) of the source domain is calculated based on the distribution difference between it and the target domain. The confidence of each source domain is fused through the DS synthesis formula to obtain a comprehensive judgment result. If the DS synthesis result is higher than the preset threshold, the corresponding alarm category and level are output.

[0041] The advantage of using the DS rule for multi-source fusion is that it can better handle possible conflicts and uncertainties between source domains. When some source domain data differs greatly from the target domain, the weight of the source domain will be weakened through the aforementioned discount coefficient adjustment, thereby reducing the impact of erroneous judgments.

[0042] In addition, in some industrial situations, alarm types may often be coupled with each other (such as the simultaneous presence of complex faults such as over-temperature and under-pressure), and if only hard decisions are made (one-time output of a single alarm category), sometimes a certain error rate will be introduced. To this end, the present invention further refines the alarm category after DS fusion, which may specifically include the following implementations: The target alarm signal is searched for neighbor samples in the high-dimensional feature space. If the neighbor samples are concentrated and have clear fault labels, they tend to be judged as a single alarm category. If the distribution of neighbors is relatively dispersed or multiple fault labels coexist, it may be judged as a composite alarm event (such as excessive temperature + valve opening by mistake; safety interlock triggering at the same time, etc.), and the system will mark it as "composite" in the alarm record and prompt the operator to conduct a comprehensive inspection of related equipment or operation links to reduce the possibility of missed reports or false alarms.

[0043] Through this trust classification model, it is possible to more accurately distinguish between single-type faults and compound faults based on the cross-domain migration classification results, while also taking into account a variety of complex alarm scenarios, thereby improving the security level of practical applications.

[0044] After completing the classification and level confirmation of the alarm event, the present invention will uniformly write information including "cross-domain fusion classification output", "alarm category", "alarm area", "timestamp", "processing status" and so on into the historical database. For multi-source heterogeneous data, the determination process of each source domain (such as source domain ID, confidence, discount coefficient) and the final DS fusion result can also be recorded for future retrospective analysis and algorithm performance evaluation.

[0045] When the user issues a query command on the graphic control system interface, the system can display the alarm record including the following key information on the visual interface: The alarm signal waveform of the target domain (after sparse reconstruction), the judgment results of the multi-source domains and their confidence, the final classification results of the DS fusion output, the judgment process of the alarm event under the trust classification model (single class or composite class), and the processing status (unprocessed, in process, or processed); If it is found in subsequent investigations that certain events have "cross-domain misjudgment" or "overestimation / underestimation of distribution differences", the association parameters between the source domain and the target domain can be adjusted in the system background, and the correction results can be written into the model configuration again, so as to continuously optimize the migration classification effect between domains.

[0046] For example, in actual deployment, multi-source heterogeneous data fusion and migration classification modules can be deployed on the edge side or cloud side according to needs. For scenarios with high real-time requirements, corresponding algorithm modules and fusion strategies can be configured on the on-site industrial control computer so that classification and linkage can be completed in the first time when an emergency alarm occurs; for distributed production devices with a large number of alarms, cross-domain fusion and offline analysis can be completed on cloud servers or data centers, which can not only reduce the burden on edge nodes, but also better aggregate alarm information from multiple plant areas, and achieve a wider range of centralized monitoring and historical big data analysis.

[0047] In the petrochemical, steel smelting, power system, pharmaceutical and other occasions with high requirements for industrial safety, the present invention, based on traditional sparse reconstruction, can not only better utilize the fault knowledge base between different domains and reduce the decline in classification accuracy caused by insufficient training samples, but also effectively identify compound alarms through DS fusion and trust classification models, reducing the false alarm and missed alarm rates at industrial sites and significantly improving the level of production safety.

[0048] In this way, the present invention can more effectively identify alarm categories and make trusted decisions in multi-source domain scenarios, and reasonably discount the distribution differences between different source domains. Finally, with the help of the trust classification model, the alarm events are further refined into single or compound categories, thereby enhancing the adaptability of the alarm management system to actual complex working conditions.

[0049] As an optional implementation, in order to deal with the misjudgment problem caused by collection errors caused by network delays or hardware failures in PLCs in industrial sites, an additional attribute domain for the PLC operating status is set up on the basis of the original cross-domain fusion classification module, and it is regarded as an independent source of evidence.

[0050] Specifically, first, in the system configuration phase, a status monitoring mechanism is established for the PLC in each domain to collect information such as its real-time operating mode, hardware self-test results, and communication error rate, and a PLC health table is maintained in the database. Subsequently, when performing cross-domain alarm fusion, a certain basic probability distribution will be assigned to each conventional source domain judgment result to indicate the support or confidence of the source domain for a certain alarm category; at the same time, the PLC attribute domain will also generate a set of basic probability distributions corresponding to the alarm category according to the current PLC health status. If it is detected that the PLC operating status is normal and the collected network communication quality is within the set threshold range, the support for each alarm category will be improved when judging the target alarm signal; if it is found that the PLC has an abnormal scanning cycle or the fault diagnosis indicates that the hardware is unstable, the support for this alarm category will be reduced or directly regarded as conflict evidence.

[0051] In this way, the original method of simple DS synthesis that only relied on the fault or alarm model output by each source domain was improved to introduce the credibility correction of the PLC attribute domain before fusion, thereby avoiding the situation where the collected data is still regarded as having the same credibility as other normal domains when the PLC hardware or network is abnormal.

[0052] During the DS synthesis process, if the PLC attribute domain indicates normal, the recognition of the alarm judgment results in the same domain can be significantly enhanced; conversely, if the credibility of the output of the PLC attribute domain is detected to be low, it means that the alarm signal obtained by the domain has large interference or distortion, and the final DS synthesis formula will discount its evidence and give it a lower weight.

[0053] After the fusion is completed, the system will not only output the synthesized alarm category and level, but also store the health information and corresponding conflicts or supporting evidence used in the calculation of the PLC attribute domain in the historical database for subsequent backtracking and evaluation.

[0054] In this way, the interference from abnormal PLCs can be better eliminated or weakened in the cross-domain fusion judgment of multiple source domains, reducing false alarms or missed alarms caused by hardware or network anomalies in a single source domain, thereby making the alarm classification results under complex working conditions more reliable and stable.

[0055] For example, in the reaction area of ​​a large petrochemical plant, three industrial-grade PLCs are deployed, including Siemens S7-400, Schneider Modicon M340 and Rockwell Allen-Bradley CompactLogix. Each PLC is connected to the host computer via industrial Ethernet. To ensure accurate monitoring of key parameters such as temperature and pressure of the reactor, each PLC uses a scan cycle of 0.1 seconds and regularly uploads data to the on-site edge computing gateway. Special signal processing software and cross-domain fusion classification modules are installed on the edge gateway. The software is developed based on a mixture of Python and C++ and can uniformly manage alarm signals from different PLCs. When the system is deployed, an independent "PLC health" record is established in the database for each PLC, including PLC_ID, scan cycle mean, communication timeout count, hardware self-test results, and reserved fault log path. The health is represented by a floating point number from 0 to 1. The closer to 1, the more stable the PLC. The database is deployed on the factory's LAN server using MySQL cluster, providing real-time insertion and historical retrieval interfaces.

[0056] During the on-site operation, when one of the PLCs detects that the temperature of the reactor exceeds the set Level II alarm threshold (for example, 250 degrees Celsius), the data stream of the temperature measurement point will be uploaded to the edge gateway along with basic fields such as timestamp and over-limit information. The signal processing software on the edge gateway first performs undersampling and sparse reconstruction on this set of temperature values, generates a reconstructed alarm signal, and then calls the cross-domain fusion classification module for preliminary analysis. The module compares the alarm signal with the existing fault models in other domains (such as heat exchange area and tank domain), and obtains the basic probability distribution for temperature alarms through the DS synthesis rule. In this process, the system additionally introduces PLC attributes as an independent source of evidence for weighting. The specific approach is: After obtaining the alarm data from the PLC, the cross-domain fusion classification module reads the current health value and interlock trigger status of the PLC from the database. If the health value is greater than 0.8 and no interlock trigger occurs, a higher support value, such as 0.6 to 0.7, will be assigned to the "high temperature alarm" category of the PLC in this alarm; if it is detected that the PLC communication timeout count is high, the hardware self-check error or the interlock trigger conflicts with the current alarm category, the support value of the corresponding alarm category will be reduced to below 0.2 or even marked as conflict evidence. This support value, together with the judgment results of other source domains, generates the final "temperature fault alarm" confidence through the Dempster synthesis formula.

[0057] Once the confidence level of the "temperature fault alarm" output by the cross-domain fusion classification module is higher than the preset threshold of 0.8, the system will confirm that the alarm is a valid fault and automatically record fields such as "Alarm Category: Temperature Fault", "Alarm Level: Level II", "PLC Attribute Domain Judgment: High Health", and "Processing Status: Unprocessed" into the historical database. At the same time, it will trigger a flashing prompt on the graphic control interface and generate an alarm log for the operator to sign and process.

[0058] If in another alarm scenario, the health of a PLC drops to 0.4 and the hardware diagnosis shows that the analog input module is abnormal, the system will identify the alarm signal as a less credible alarm source based on the evidence output by the attribute domain, and the DS synthesis formula will correspondingly reduce the final support for this alarm type. If the fusion result is lower than the threshold value, the system will only mark it as "pending review" in the alarm list and prompt the operator to prioritize the PLC or confirm whether its communication status is normal. The attribute domain judgment result will also be stored in the database as an independent "fusion auxiliary record" so that the operation and maintenance personnel can see detailed health values, communication timeout counts, and final discount coefficients when querying the PLC operating status when the specific alarm occurs, thereby providing reliable data support for alarm analysis and troubleshooting.

[0059] Through the above method, multiple PLCs on site can stably provide reliable alarm data under normal circumstances. Once a hardware failure or network jitter occurs, the alarm evidence belonging to the PLC will be automatically discounted during DS synthesis, and will not cause significant interference to the overall alarm judgment. Using PLC attributes as an independent source of evidence can significantly reduce the risk of false alarms or missed alarms caused by single-point PLC anomalies in multi-source domain fusion scenarios, ensuring the safety interlocking of petrochemical equipment and the stability of process operation.

[0060] As an optional implementation, in some complex industrial scenarios, especially when the PLC sampling cycles are not completely consistent or there are delays in the field network, the alarm signals uploaded by each source domain may have timing misalignment, missing point filling or large differences in sampling rates, resulting in the inability to accurately reflect the true differences between the source domain and the target domain when directly comparing the feature distributions of different domains.

[0061] Therefore, a divergence calculation method integrating time series alignment and local weighting strategy is adopted to solve the classification inaccuracy problem caused by time series mismatch in the graphics control system by dynamically aligning the data before divergence calculation.

[0062] In the specific implementation, the reconstructed alarm signal of each source domain is first time-stamped and standardized in the cross-domain fusion classification module, and the feature vector is resampled or interpolated according to an adjustable time step, so that the sampling time series of the source domain and the target domain are as consistent as possible on the same time axis. If there are abnormal PLC reports or communication delay records at certain sampling points, the data of this section is weighted interpolated or eliminated at this stage. Subsequently, a local weighted calculation method is introduced on the realigned feature vector: a higher weight is given to the time period or feature peak section that is closer to the alarm trigger point of the target domain; for the section far away from the critical time period or with a large interpolation error, an attenuation coefficient is used to reduce its influence. By combining this local weighting with time alignment, the signal similarity between the source domain and the target domain at key moments can be more accurately measured while retaining the overall distribution comparison, thereby obtaining a corrected "weighted divergence value".

[0063] At the implementation level, it can be transformed based on the MMD (Maximum Mean Discrepancy) framework: when calculating the kernel function, not only the Euclidean distance or kernel mapping similarity of the feature points between the source domain and the target domain is considered, but also the timing difference and local weight factor after time alignment are introduced.

[0064] Specifically, when calculating a feature point pair, its timestamp is compared with the timestamp of the target domain feature point. If the time difference between the two is small and close to the alarm peak position, the kernel function output is improved; if the time difference between the two is large or there are many missing data segments in the source domain, the kernel output contribution of the point pair is reduced. After completing the summation of the kernel function, the obtained MMD value can better reflect the real domain difference. Finally, when the system performs cross-domain fusion classification, it generates a new reliability or discount factor based on this weighted divergence value, assigning higher credibility to source domains with lower divergence values ​​(indicating that the two domains are very similar in the critical period after alignment), and vice versa, increasing the discount factor to weaken its impact.

[0065] In the specific implementation, it is necessary to add a "time series alignment" strategy configuration item in the database or configuration file to record the sampling period and network delay baseline of each PLC, so as to facilitate dynamic compensation of the feature sequence according to the real-time measured communication delay after the alarm is triggered. The time alignment offset and local weighted distribution of each divergence calculation process can also be saved in the historical database, so that the operation and maintenance personnel can clearly see the "alignment error between this domain and the target domain within a certain time range" and "in which sampling segment the interpolation and elimination processing is performed" and other information when looking back, so as to better understand the final formation process of the divergence value. If a PLC has serious data anomalies at a certain stage, the system can also find a large number of high interpolation or high elimination records based on the time series alignment results, and then further reduce the support of the domain through the PLC attribute domain evidence source to avoid misleading the final alarm fusion decision.

[0066] In this way, when the graphics control system makes cross-domain fusion judgments in multiple source domains, it can not only quantify simple distribution differences, but also make more delicate dynamic corrections for inconsistent sampling cycles, network jitter, and local PLC anomalies that are unique to industrial sites, reducing misjudgments caused by timing misalignment or data loss. When combined with the aforementioned solution for the PLC attribute domain, this improved divergence calculation can also further enhance the recognition and discount of abnormal data, and improve the stability and accuracy of the graphics control system's alarm classification under complex working conditions.

[0067] As an optional implementation, it also includes: In the sparse reconstruction algorithm, the target PLC is set as the PLC that generates a compressed alarm signal and triggers an alarm state through the graphic control system; Setting an association relationship table for the signal association between each PLC in the historical database to store association parameters obtained based on similarity or mutual information; A multi-source dictionary management module is configured in the sparse reconstruction algorithm to load the historical waveform basis functions of the target PLC and the associated PLC; In response to the health value of the target PLC being less than a preset threshold, the multi-source dictionary management module references a corresponding pattern from the waveform basis function of the associated PLC and reduces the weight of the target PLC local basis function in the reconstruction calculation to generate the reconstruction alarm signal; In response to the health value of the target PLC being greater than or equal to the preset threshold, the multi-source dictionary management module performs reconstruction calculation using the local basis function of the target PLC and outputs the reconstruction alarm signal; After the reconstruction alarm signal is generated, the dictionary call information of the multi-source dictionary management module is written into the historical database together with the health records of the target PLC and the associated PLC.

[0068] The compressed sensing and sparse reconstruction described above mainly rely on a single signal or a historical feature library of the same domain (such as each measuring point under a PLC) to complete the reconstruction. However, in the actual industrial field graphics control system, if the operating status of a PLC is unstable or the measurement accuracy of some sensors decreases, simply relying on its local historical data for reconstruction may cause distortion of the reconstruction result or insufficient recognition of sudden fault characteristics. To this end, on the basis of the original compression and sparse reconstruction, combined with the relationship between multiple PLCs and the health information of each PLC, a dynamic multi-source dictionary or cross-PLC collaborative reconstruction strategy can be designed, so that when the credibility of a certain PLC data decreases, it can still maintain a high reconstruction quality and improve the stability of alarm recognition.

[0069] In the specific implementation, during the system configuration phase, in addition to establishing a health record table for each PLC (including scan cycle average, communication timeout count, hardware self-test results and fault log, etc.), a PLC association table will be established in the database for the signal correlation between each PLC. This association can be calculated based on the time series similarity, covariance or mutual information of historical data, and stored in the form of a matrix or list.

[0070] For example, in a petrochemical production plant, multiple PLCs monitor similar process units (such as adjacent reactors and heat exchangers). Since their temperatures or pressures change synchronously with the process, they often show a high correlation coefficient. If some PLCs are far apart in terms of geographical location or process flow, their signal correlation may be low.

[0071] When a PLC detects an alarm state and uploads a compressed alarm signal, the system will additionally query the current health value of the PLC and its association with other PLCs before executing the step of "reconstructing the compressed alarm signal through a sparse reconstruction algorithm". If the health value is low (such as lower than 0.5) and the communication timeout count increases, it means that the PLC may have data reliability issues, and the system will increase the reference weight of historical samples provided by other PLCs with high correlation and normal health (such as greater than 0.8).

[0072] The specific approach is that when establishing the dictionary library required for sparse reconstruction, in addition to loading the historical pattern basis functions of this PLC, a portion of typical fault / normal sample waveforms from other PLCs with high correlation will be selected to expand the dictionary or increase the priority of the corresponding basis functions. In this way, if the local data of the PLC is incomplete due to failure or interference, the reconstruction algorithm can also obtain useful patterns from other PLC data with higher health and similarity, so that the reconstructed alarm signal is closer to the actual situation in terms of amplitude and timing.

[0073] At the software implementation level, a dynamic dictionary management module can be maintained on the edge computing node or industrial control server. During reconstruction, this module first reads the PLC association table and the target PLC health information. If the health of the target PLC is higher than a certain threshold (such as 0.8), the historical waveform basis function of the PLC itself is mainly used; if the health is between 0.5 and 0.8, the waveform basis function of the associated PLC will be moderately introduced to reduce the noise or distortion that may be caused by local data; if the health is lower than 0.5 and communication failures occur frequently, the waveform basis function from the high-health PLC will be used as the main reference, and a discount coefficient will be applied to the basis function of this PLC (reducing the weight in the dictionary search process) to maximize the accuracy of the reconstruction result. This process can be achieved through weighted sparse coding or multi-source dictionary fusion, for example, in the K-SVD dictionary update or orthogonal matching pursuit (OMP) process, a higher priority is assigned to the associated PLC basis function.

[0074] At the hardware deployment level, sufficient computing and storage resources can be configured on edge computing nodes (such as industrial PCs or dedicated AI gateways) for online multi-source dictionary learning and sparse reconstruction. If the system is large and needs to process data from multiple PLCs at the same time, distributed dictionary management and reconstruction algorithms can be used on the server or cloud, and the real-time status (including PLC health, associations, etc.) can be read from the database cache (such as Redis or memory tables) to reduce the latency of relational database queries. For scenarios that require extremely high real-time performance on site, the minimized dictionary can be retained locally and quickly matched, while complex multi-source updates or global dictionary training can be placed on the back-end server for periodic execution.

[0075] In this way, when a certain PLC is abnormal, the alarm signal compression and reconstruction process no longer relies solely on the historical basis function of the PLC, but uses the health data of the highly correlated PLC to assist in correction, thereby achieving a more accurate restoration of the fault segment or distorted segment. Compared with the traditional single PLC dictionary reconstruction, this solution can better alleviate the reconstruction deviation caused by single-point failures, and improve the overall alarm reconstruction accuracy in situations such as petrochemical or power systems where multiple PLCs are interrelated, thereby making the alarm judgment obtained in the cross-domain fusion classification stage more reliable and more in line with actual process requirements.

[0076] For example, when a PLC has a long-term drift or noise surge in its temperature channel due to a hardware failure, other PLCs or sensors with a high correlation with the PLC can still use their normally collected data to supplement the reference waveforms under similar process conditions for the target PLC, making the sparse reconstruction of the compressed signal more accurate in capturing peaks and trends; conversely, when the system detects that the health of the associated PLC itself is also declining, it will reduce its reliance on its data to avoid mistaking abnormal values ​​for baseline reconstruction. Thus, under the mode of multiple redundancy and cross-PLC collaboration in industrial sites, it is possible to obtain high-quality alarm signal reconstruction results under the premise of differences in equipment health status and multi-source heterogeneous signals, solving the problem of traditional single-source sparse reconstruction being easily distorted or missing data when the PLC is abnormal.

[0077] See also Figure 4 , Figure 4 A schematic diagram of a health record maintenance process using a vibration unit provided in an embodiment of the present application includes steps S201 to S203, wherein: S201: installing a vibration trigger for the monitoring object where the target PLC is located, and establishing a vibration reference table for each vibration mode in the historical database to store a preset excitation frequency and a theoretical response curve; S202: When the graphics control system is in a non-critical operation period, an active vibration instruction is sent to the vibration trigger, and sampling data of the target PLC in the vibration period is obtained; S203: After comparing the sampled data with the theoretical response curve, if the deviation between the two is greater than a threshold, an active vibration detection abnormality is marked in a health record table corresponding to the target PLC.

[0078] As an optional implementation, a vibration basis function library may be provided in the sparse reconstruction algorithm to store standard waveform basis functions corresponding to each excitation mode in the vibration reference table; In response to an active vibration detection abnormality flag of a target PLC being written, applying a discount coefficient to a local basis function used by the target PLC in a reconstruction calculation, and increasing the priority of the associated PLC basis function in a multi-source dictionary management module; After the reconstruction is completed, the reference information of the vibration base function library and the active vibration detection abnormality mark of the target PLC are written into the historical database.

[0079] As an optional implementation, a vibration excitation cycle configuration item may be set in the database or configuration file to periodically trigger the vibration trigger and collect vibration response data of the target PLC; In response to detecting that the vibration response of the target PLC deviates from the theoretical response curve by an amplitude or frequency exceeding a preset range for multiple times during the excitation cycle, automatically reducing the health value of the target PLC to below a predetermined threshold; After the health value decreases, when cross-domain reconstruction is performed on the compressed alarm signal uploaded by the target PLC based on the sparse reconstruction algorithm, other PLC basis functions whose health values ​​are greater than or equal to a predetermined threshold are called to compensate for the data distortion of the target PLC, and the final reconstruction result is written into the historical database.

[0080] Based on the existing solutions based on multi-source dictionaries and cross-PLC collaborative reconstruction, the present application further uses a mechanism of active vibration triggering and signal feedback acquisition to provide an additional reference signal for the PLC, so as to improve the reconstructed alarm signal generated by reconstructing the compressed alarm signal through a sparse reconstruction algorithm, thereby improving the accuracy and calibratability of the data collected by the sensor or PLC under certain special working conditions.

[0081] In the above scheme, the PLC triggers alarms and reports data only based on process variables monitored by sensors (such as temperature, pressure, liquid level, flow, etc.). When potential faults in sensors or PLCs (such as decreased sensitivity, reference drift, range damage) are not discovered in time, the uploaded data will affect the quality of subsequent compressed sampling and sparse reconstruction, which may lead to alarm errors.

[0082] To this end, the present application deploys a controllable vibration unit near the PLC (or on the mechanical structure of the monitored object). The vibration unit is not a necessary part of the production process, but is specifically used to actively generate vibration excitation at a specific moment and allow the sensor to capture the corresponding feedback signal of this vibration. By comparing the measured vibration signal with the expected theoretical waveform or historical reference, the health or deviation of the sensor and its corresponding PLC channel can be evaluated; once it is found that the characteristic response of the sensor is too far from the standard response, the data weight of the PLC can be discounted or corrected in the subsequent reconstruction process, or the matching process of the dictionary library can be calibrated.

[0083] At the data level, the system needs to add an "active vibration reference table" for each PLC in the database. The table contains the vibration mode ID, excitation frequency, excitation amplitude, corresponding theoretical response curve, and applicable measurement point range. According to the rigidity, material properties or structural form of different equipment on site, multiple sets of vibration excitation schemes can be configured, such as different vibration frequencies for thicker and thinner pipes, or different sensing frequency bands for temperature sensors and pressure transmitters. Once a vibration mode ID is selected, the PLC will trigger the vibration unit to excite at a specified time according to the corresponding instructions, and collect the entire vibration response data from the sensor end.

[0084] At the hardware level, a vibration trigger and drive unit can be installed near the field equipment or PLC cabinet. The vibration trigger can use a piezoelectric ceramic actuator, an electromagnetic vibrator or other industrial vibration device, which can generate mechanical vibration of a certain frequency, amplitude or waveform after the command is issued. When the PLC executes the "active vibration test" command, it will first record the start time of the vibration, then collect the sensor signal at a high frequency in a preset time window (for example, within a few seconds), and make a preliminary comparison between this vibration response and the reference curve. If the comparison result shows that the deviation value exceeds a certain threshold, such as "the vibration peak drop is greater than 20%" or "the main resonant frequency offset exceeds a certain standard", it means that the sensor or PLC input module may be attenuated or miscalibrated. At this time, the PLC will write an "active vibration detection abnormality" flag to the health table, and the edge computing or server will moderately reduce the weight of the data provided by the PLC when performing compression and reconstruction in the subsequent execution.

[0085] At the software implementation level, in addition to the above health indicators, the vibration response record can also be included in the sparsely reconstructed dictionary library.

[0086] The specific approach is: when configuring the system, a corresponding "standard waveform basis function" is established for each vibration mode ID, and after the PLC completes the vibration test, the actual collected "measurement waveform" is also written into the "vibration comparison record table". If the actual measurement is not much different from the standard, it can be considered that the current state of the measurement point is good; if the difference is significant, the historical basis function under the PLC is updated or marked as "low confidence" to avoid the subsequent reconstruction algorithm relying heavily on this part of the inaccurate basis function. A further optimization is that this "active vibration" can be initiated periodically under normal working conditions, so as to perform online detection of the drift of the sensor and PLC input channel, forming a set of "dynamic calibration" processes; once the inaccuracy is detected to accumulate to a certain extent, the operation and maintenance personnel can decide whether to perform on-site maintenance or replace the sensor.

[0087] At the algorithm processing level, combined with the above-mentioned solutions on cross-PLC association and health discount, when the target PLC is marked as having low health due to abnormal vibration test results, the system can further reduce the weight of the local historical waveform basis function extracted from the PLC during reconstruction; at the same time, if other high-health PLCs have performed well historically under the same or similar vibration modes, more reference modes can be supplemented from the basis functions of these PLCs to ensure that the signal can still be accurately reconstructed when an alarm state occurs.

[0088] For the use scenario of multi-source dictionary, if the vibration pattern ID is the same and the device structure is similar, the vibration waveform features can be shared among multiple PLCs, so that when a PLC performs poorly, the vibration features of other healthy PLCs can make up for it. The reconstructed alarm signal obtained in this way is closer to the actual fault state in terms of features such as amplitude and phase, reducing distortion caused by sensor drift or hardware failure.

[0089] For example, an electromagnetic vibrator is installed in a certain pipeline section. The system issues a trigger command during non-critical operation periods to make the vibrator vibrate at a frequency of 150 Hz for 2 seconds. The PLC captures the response curve through a temperature sensor or an acceleration sensor and stores it in the "Vibration Comparison Record Table". After comparing the curve with the standard curve, if the peak deviation is small and the main harmonic frequency is consistent, the health of the PLC is not affected; if there are abnormalities such as a peak deviation of >30% and a main harmonic frequency misalignment of more than 0.5 times, the health of the PLC is lowered to 0.6 and the database is updated. If the PLC subsequently has a real fault alarm on the same day, the uploaded compressed alarm signal will process the local basis function in a discounted manner during sparse reconstruction, and will refer to the basis functions of other related PLCs more frequently, avoiding using sensor data with severe drift as a reliable basis, thereby reducing the probability of false alarms or missed alarms.

[0090] In this way, the "contrast waveform" collected by the active vibration trigger can directly or indirectly participate in the PLC health assessment and the optimization of the sparse reconstruction algorithm, fundamentally reducing the alarm data quality problems caused by sensor inaccuracy or hardware failure. For large equipment groups or remote distributed PLCs, it can significantly improve the efficiency of predicting faults and online diagnosis, so that the graphics control system can still maintain high-quality reconstruction and accurate classification of alarm signals under complex working conditions. Especially in situations where high-precision monitoring of vibration, noise or other dynamic characteristics is required (such as rotating machinery, vibrating screening devices, etc.), this solution can effectively solve the problem that it is difficult for passive observation mode to detect sensor accuracy attenuation in time, allowing operation and maintenance personnel to more actively grasp the actual status and correct the basis function library of compressed sensing and reconstruction algorithms in real time.

[0091] It should be noted that in the application process of the solution of the present invention, the required software and hardware configurations can be implemented based on the PLC control systems and data acquisition devices commonly used in industrial sites, and can be elastically expanded in combination with existing edge computing or cloud server deployment methods. Whether it is a small-scale demonstration device within a single plant or a large-scale distributed industrial system across regions and plants, it can be adapted to a variety of process flows and actual production needs by flexibly configuring functional modules such as "vibration trigger strategy", "multi-source dictionary management module", and "PLC health record table". Through seamless integration with on-site process control logic, the solution can provide operators or system maintenance personnel with a more detailed alarm data basis and accurate fault identification results without significantly increasing the burden of production operations, thereby achieving an effective balance between smooth and safe operation and rapid fault diagnosis.

[0092] The solution described in the present invention effectively combines multiple methods such as compressed sensing and sparse reconstruction, DS fusion reasoning, multi-source PLC association and active vibration calibration at the technical level, and connects the acquisition, reconstruction, fusion and health assessment of alarm data into a set of systematic processes. Through the optimization design of industrial sensor characteristics in vibration triggers and health management mechanisms, the present invention not only reduces the probability of false alarms or missed alarms caused by sensor failures, PLC anomalies or network delays, but also makes the results of compressed reconstruction closer to the dynamic changes of the actual production process; coupled with the coordinated use of technical means such as time series alignment and local weighted divergence in the multi-source fusion stage, the entire alarm management process has higher robustness and scalability, and can be widely used in occasions with high requirements for alarm accuracy and real-time performance.

[0093] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0094] The preferred embodiments of the present invention disclosed above are only used to help explain the present application. The preferred embodiments do not describe all the details in detail, nor do they limit the specific implementation methods of the present application. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present application, so that those skilled in the art can understand and use the present application well. The present application is limited only by the claims and their full scope and equivalents.

Claims

1. The industrial safety alarm event recording and tracing method based on the graphic control system is characterized by: include: Collecting an original alarm signal, and performing undersampling processing on the original alarm signal to generate a compressed alarm signal; Reconstructing the compressed alarm signal by using a sparse reconstruction algorithm to generate a reconstructed alarm signal; According to preset area division rules and alarm type classification rules, the reconstructed alarm signal is divided into alarm events of different alarm levels and different alarm areas; According to the divided alarm events, the alarm level, area information, occurrence time and processing status of the alarm event are automatically recorded in the historical database through the pre-associated PLC data points to form the alarm event history record; Receive a user's query instruction, extract corresponding alarm event history records from the history database according to the query instruction, and perform visual tracing display on the graphics control system interface.

2. The industrial safety alarm event recording and tracing method based on the graphic control system according to claim 1 is characterized in that: Also includes: When configuring the graphics control system, a domain partition library is created, and a domain identifier is assigned to each heterogeneous data source in the domain partition library, and a mapping relationship is established between the domain identifier and the corresponding sensor list, sampling frequency parameter, and production unit or plant area identifier; In response to a reconstructed alarm signal within a domain exceeding a preset threshold, the reconstructed alarm signal and the historical reference data corresponding to the domain are sent to a cross-domain fusion classification module to obtain a preliminary alarm classification result for the domain. After obtaining the classification result, the domain identifier and the corresponding preliminary alarm classification result are written into a historical database.

3. The method for recording and tracing industrial safety alarm events based on a graphic control system according to claim 2 is characterized in that: Also includes: A distribution difference measurement unit is set in the cross-domain fusion classification module, and the distribution difference measurement unit calculates the distribution difference degree based on the characteristics of the reconstructed alarm signals collected from each source domain and the target domain in the domain division library and based on a preset divergence calculation method; The cross-domain fusion classification module is also provided with a discount coefficient calculation unit for generating reliability or discount coefficient according to the distribution difference; After completing the calculation of distribution difference and discount coefficient, the alarm judgment results of each source domain are weighted and fused, and the comprehensive alarm category and alarm level are output. A comprehensive judgment result is obtained through a preset synthesis formula. When the judgment result is higher than the preset fusion threshold, the comprehensive alarm category and alarm level are written into the historical database.

4. The method for recording and tracing industrial safety alarm events based on a graphic control system according to claim 3 is characterized in that: Also includes: A trust classification model is configured in the cross-domain fusion classification module, wherein the trust classification model includes a neighbor search unit and a composite class determination unit, wherein the neighbor search unit is used to identify the concentration of fault labels based on the neighborhood distribution of the reconstructed alarm signal in the high-dimensional feature space; The comprehensive alarm category obtained in response to the preset synthesis formula includes multiple fault labels, which triggers the composite category determination unit to identify the alarm event as a composite alarm, and records the source domain information, alarm category, alarm area and timestamp of the composite alarm to the historical database, so that the graphics control system interface can trace the source and display the composite alarm event information when receiving a user query instruction.

5. The method for recording and tracing industrial safety alarm events based on a graphic control system according to claim 4 is characterized in that: Also includes: Setting a PLC attribute domain evidence source in the cross-domain fusion classification module and storing a health record table for each PLC in a historical database; The health record table includes the scan cycle average, communication timeout count and hardware self-test result, which are used to indicate the operating status of the PLC; Before calculating the distribution difference of the reconstructed alarm signal and fusing it with the synthesis formula, querying the health record table of the PLC attribute domain evidence source and generating the corresponding attribute domain confidence; In response to the health record indicating that the PLC is in a normal operating state, increasing support for the alarm category; In response to the health record table indicating that the PLC communication timeout or hardware failure is detected, the support or confidence of the alarm category is reduced, and the fusion result corrected by the PLC attribute domain evidence source is written into the historical database.

6. The method for recording and tracing industrial safety alarm events based on a graphic control system according to claim 5 is characterized in that: Also includes: In the sparse reconstruction algorithm, the target PLC is set as the PLC that generates a compressed alarm signal and triggers an alarm state through the graphic control system; Setting an association relationship table for the signal association between each PLC in the historical database to store association parameters obtained based on similarity or mutual information; A multi-source dictionary management module is configured in the sparse reconstruction algorithm to load the historical waveform basis functions of the target PLC and the associated PLC; In response to the health value of the target PLC being less than a preset threshold, the multi-source dictionary management module references a corresponding pattern from the waveform basis function of the associated PLC and reduces the weight of the target PLC local basis function in the reconstruction calculation to generate the reconstruction alarm signal; In response to the health value of the target PLC being greater than or equal to the preset threshold, the multi-source dictionary management module performs reconstruction calculation using the local basis function of the target PLC and outputs the reconstruction alarm signal; After the reconstruction alarm signal is generated, the dictionary call information of the multi-source dictionary management module is written into the historical database together with the health records of the target PLC and the associated PLC.

7. The method for recording and tracing industrial safety alarm events based on a graphic control system according to claim 6 is characterized in that: Also includes: Install a vibration trigger for the monitoring object where the target PLC is located, and establish a vibration reference table for each vibration mode in the historical database to store the preset excitation frequency and theoretical response curve; When the graphics control system is in a non-critical operation period, an active vibration instruction is sent to the vibration trigger, and sampling data of the target PLC in the vibration period is obtained; After comparing the sampled data with the theoretical response curve, if the deviation between the two is greater than a threshold, an active vibration detection abnormality is marked in the health record table corresponding to the target PLC.

8. The method for recording and tracing industrial safety alarm events based on a graphic control system according to claim 7 is characterized in that: Also includes: A vibration basis function library is provided in the sparse reconstruction algorithm for storing standard waveform basis functions corresponding to each excitation mode in the vibration reference table; In response to an active vibration detection abnormality flag of a target PLC being written, applying a discount coefficient to a local basis function used by the target PLC in a reconstruction calculation, and increasing the priority of the associated PLC basis function in a multi-source dictionary management module; After the reconstruction is completed, the reference information of the vibration base function library and the active vibration detection abnormality mark of the target PLC are written into the historical database.

9. The method for recording and tracing industrial safety alarm events based on a graphic control system according to claim 8 is characterized in that: Also includes: Setting a vibration excitation cycle configuration item in the database or configuration file to periodically trigger the vibration trigger and collect vibration response data of the target PLC; In response to detecting that the vibration response of the target PLC deviates from the theoretical response curve by an amplitude or frequency exceeding a preset range for multiple times during the excitation cycle, automatically reducing the health value of the target PLC to below a predetermined threshold; After the health value decreases, when cross-domain reconstruction is performed on the compressed alarm signal uploaded by the target PLC based on the sparse reconstruction algorithm, other PLC basis functions whose health values ​​are greater than or equal to a predetermined threshold are called to compensate for the data distortion of the target PLC, and the final reconstruction result is written into the historical database.

Images