A device monitoring method, system, device, and medium based on AI and Industrial Internet of Things

Abstract

This invention provides a device monitoring method, system, device, and medium based on AI and Industrial Internet of Things (IIoT), belonging to the field of device operation monitoring technology. The method includes: acquiring device monitoring operations; triggering a device monitoring process; acquiring key monitoring datasets based on the monitoring process; querying a preset key data analysis rule table based on the key monitoring data to obtain device data parsing parameters for the key monitoring data; querying a preset monitoring task sequence to obtain monitoring process tags; calling a device operation status analysis process based on the monitoring process tags; analyzing the monitored device through the device operation status analysis process to obtain monitoring target data; and parsing the key monitoring data based on the monitoring target data to obtain the target device monitoring result. This invention achieves accurate monitoring of the operating status of IIoT devices. It can promptly capture abnormal situations and key information during device operation, improving the efficiency and reliability of monitoring work.

Description

Technical Field

[0001] This invention belongs to the field of equipment operation monitoring technology, specifically relating to an equipment monitoring method, system, equipment, and medium based on AI and industrial IoT. Background Technology

[0002] Industrial Internet of Things (IIoT) systems typically employ a periodic data collection model. Industrial gateways located at the network edge collect operational data from various industrial devices at preset frequencies and upload this raw data indiscriminately to a cloud platform. The cloud platform undertakes the core functions of data storage, analysis, and application, such as monitoring equipment operating status and providing fault warnings.

[0003] In the typical application scenario of equipment operation status monitoring, most of the data uploaded to the cloud actually reflects the normal operating status of the equipment. In related technologies, equipment data parsing rules are often not adjusted based on equipment load status, operating mode, abnormal conditions, and maintenance time. This leads to a mismatch between parsing parameters and actual data acquisition accuracy under different operating scenarios, resulting in significant deviations in parsing results. In monitoring methods, it is difficult to adjust the parsing order according to different monitoring objectives and the criticality of data, thus failing to improve monitoring effectiveness, affecting monitoring efficiency and the timely detection of critical anomalies. It is also difficult to capture complex abnormal patterns in equipment operation, easily leading to missed or false diagnoses. Finally, it cannot provide comprehensive decision-making basis for equipment maintenance and fault diagnosis. Summary of the Invention

[0004] This invention provides a device monitoring method based on AI and the Industrial Internet of Things (IIoT), enabling accurate monitoring of the operational status of IIoT devices. It can promptly capture abnormal situations and key information during device operation, improving the efficiency and reliability of monitoring work.

[0005] The methods include:

[0006] S101: Obtain device monitoring operations;

[0007] S102: Trigger the device monitoring process based on the device monitoring operation, and obtain the key monitoring dataset based on the device monitoring process; the key monitoring dataset includes key monitoring data.

[0008] S103: Query the preset key data analysis rule table based on the key monitoring data to obtain the device data parsing parameters of the key monitoring data;

[0009] S104: Query the preset monitoring task sequence based on key monitoring data to obtain monitoring process tags;

[0010] S105: Invoke the device running status analysis process based on the monitoring process label;

[0011] S106: Analyze the monitored equipment through the equipment operation status analysis process to obtain the monitoring target data;

[0012] S107: Analyze the key monitoring data based on the monitoring target data to obtain the monitoring results of the target device.

[0013] It should be further explained that step S102 specifically includes:

[0014] Determine the trigger condition type for device monitoring operations. Trigger condition types include: periodic triggering, event-driven triggering, or manual active triggering.

[0015] The device monitoring process is matched according to the trigger condition type, and the matching process is based on the preset functional mapping relationship between the trigger type and the monitoring process;

[0016] Extract sensor data source information related to the current monitoring process;

[0017] Based on the data source information, the data acquisition interface is called to collect real-time operating parameters during the device operation process, and the data is encapsulated according to the preset data structure to form a key monitoring dataset containing key monitoring data.

[0018] It should be further explained that step S107 specifically includes:

[0019] Determine the parsing order of key monitoring data. The parsing order is determined based on the application scenario and the criticality of the monitoring target data.

[0020] Match the model tags corresponding to the monitored target data from the preset model tag library;

[0021] According to the established parsing order, the matching device monitoring model is invoked to parse the key monitoring data step by step. The parsing process includes feature extraction and preliminary judgment of the data, and the model parameters are adjusted according to the real-time data characteristics during the parsing process.

[0022] The device features obtained through step-by-step parsing are aggregated according to business requirements. The aggregation method includes time and logical dimensions, ultimately forming a target device monitoring result that covers the device's operating status.

[0023] It should be further explained that step S107 specifically includes:

[0024] The method for determining the parsing order of key monitoring data is as follows:

[0025] MI ( X , Y )= α ⋅ MI ( X ,Y )+ β ⋅ H ( Y | X )+ c ⋅Corr( X , Y )

[0026] in, X To monitor key data feature sets, Y To monitor the target data feature set, H ( Y | X ) is the conditional entropy, Corr( X , Y ) represents the Pearson correlation coefficient. α , β , c corresponding These are the weighting coefficients;

[0027] The device monitoring model parses data based on a threshold function, where the threshold function is:

[0028] i ( t )= i 0+ k ⋅exp(−2 s 2( t - t 0)2)

[0029] in, i 0 is the preset base threshold. k For adjustment coefficients, t For the current time, t 0 represents a period of high frequency of historical equipment anomalies. s The standard deviation of the time window;

[0030] Device features are aggregated using the information entropy formula, which is:

[0031] H ( D )=−∑ i =1 npi ⋅log( pi )+ l ⋅max(0, s 2∑ i =1 m ( xi - m )2− t )

[0032] in, D For the set of device features, pi For the first iThe probability of class data appearing l The regularization coefficient is . m For data dimensions, xi For the first i Dimensional data values, m The mean, s 2 represents the variance. t This is the tolerance threshold for abnormal fluctuations.

[0033] It should be further explained that step S103 specifically includes:

[0034] Extract feature tags from key monitoring data. The feature tags include data type, time attribute, device association attribute and anomaly risk level, forming a set of query conditions.

[0035] Filter the preset key data analysis rule table based on feature tags;

[0036] The retained candidate rules are conflict resolved. The conflict resolution is based on the confidence, timeliness and coverage of the rules, and the rule with the highest confidence, the strongest timeliness and the best coverage is selected as the target parsing rule.

[0037] Based on the target parsing rules, the parsing parameters defined in the rules are extracted, and the parsing parameters are corrected according to the actual acquisition accuracy of the key monitoring data, so as to finally obtain the device data parsing parameters that are adapted to the current key monitoring data.

[0038] It should be further explained that the steps for conflict resolution of the retained candidate rules specifically include:

[0039] Construct a multi-dimensional rule resolution indicator set, forming an indicator set that includes confidence level C, timeliness T, coverage S, interpretability E, and cross-system compatibility I:

[0040] Dynamic weight values ​​are assigned to each dimension based on device type and current task objective;

[0041] Based on the weighted index value, index value = C×Wc+T×Wt+S×Ws+E×We+I×Wi, calculate the priority of each candidate rule; at the same time, prioritize the task target with the higher comprehensive score.

[0042] The accuracy of rule parsing in the current scenario is verified by simulating working condition data.

[0043] If the verification result deviates from the actual expectation by more than a preset threshold, the coverage of the rule will be adjusted and the candidate will re-enter the candidate pool.

[0044] If the rule fails to meet the standard after multiple verifications, it will be downgraded to a backup rule. Finally, the rule that passes verification and has the highest overall score will be selected as the target parsing rule.

[0045] It should be further explained that step S106 specifically includes:

[0046] Data fusion preprocessing is used to align and integrate data from different channels and formats.

[0047] Based on the current operating condition and historical context information of the equipment, dynamic features closely related to the equipment status are extracted from the fused dataset, making the features adaptable to different scenarios.

[0048] Using a pre-set status assessment rule table and an adaptive assessment model, combined with extracted dynamic features, the assessment threshold is dynamically adjusted according to real-time operating conditions. The equipment operating status is graded and assessed by integrating the correlation of multiple features, and status conclusions and key influencing factors are obtained.

[0049] The status assessment results are correlated with the original data, features, and assessment criteria to generate structured monitoring target data.

[0050] This application also provides an equipment monitoring system based on AI and industrial IoT, the system including: a control unit, a data acquisition unit, a data analysis unit, and a data reporting unit;

[0051] The data acquisition unit is used to acquire sensor data;

[0052] The control unit is used to coordinate the work of each unit and acquire equipment monitoring data.

[0053] The data analysis unit is used to trigger the device monitoring process based on device monitoring operations, and to obtain the key monitoring dataset based on the device monitoring process; the key monitoring dataset includes key monitoring data.

[0054] The system queries the preset key data analysis rule table based on the key monitoring data to obtain the device data parsing parameters of the key monitoring data; it queries the preset monitoring task sequence based on the key monitoring data to obtain the monitoring process label; it calls the device operation status analysis process based on the monitoring process label; it analyzes the monitored device through the device operation status analysis process to obtain the monitoring target data; and it parses the key monitoring data based on the monitoring target data to obtain the monitoring results of the target device.

[0055] The data reporting unit communicates with the cloud server and reports the monitoring results of the target device to the cloud server via the network.

[0056] According to another embodiment of this application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the AI- and Industrial Internet of Things-based device monitoring method.

[0057] According to another embodiment of this application, a storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the AI- and Industrial Internet of Things-based device monitoring method.

[0058] As can be seen from the above technical solutions, the present invention has the following advantages:

[0059] This application discloses an AI- and Industrial IoT-based equipment monitoring method that automatically initiates equipment monitoring by triggering monitoring processes and acquiring key datasets through operational actions. This ensures the real-time and targeted nature of data collection and enables timely responses to monitoring needs. Parsing parameters are obtained based on a key monitoring data query rule table. Parsing rules can be adapted to data characteristics, and rules are filtered and conflicts resolved by combining equipment operating conditions and historical context, making the parsing parameters more consistent with actual operating scenarios and improving the accuracy and adaptability of data parsing. By querying preset task sequences to obtain monitoring process tags and calling analysis processes, standardized and process-oriented management of monitoring tasks is achieved. This allows different monitoring targets to be quickly matched with corresponding analysis logic, improving the efficiency and operability of the monitoring process. The target data is obtained and parsed using the analysis process. Through determining the parsing order, hierarchical parsing, and multi-dimensional aggregation, combined with information theory and dynamic thresholding methods, the analysis of equipment operating status is realized, improving the comprehensiveness and reliability of monitoring results. Attached Figure Description

[0060] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 This is a flowchart of a device monitoring method based on AI and Industrial IoT.

[0062] Figure 2 This is a schematic diagram of an equipment monitoring system based on AI and the Industrial Internet of Things.

[0063] Figure 3 This is a schematic diagram of an electronic device. Detailed Implementation

[0064] The device monitoring method based on AI and industrial IoT involved in this application can be applied to specific electronic devices, which have control units, data analysis units, and data reporting units.

[0065] The system includes a data acquisition unit that acquires sensor data. This unit features serial and Ethernet (ETH) interfaces, as well as various other interfaces such as USB, SPI, and I2C, to accommodate sensor devices with different interface types. Simultaneously, the data acquisition unit implements verification and error correction mechanisms to ensure data accuracy and reliability.

[0066] Sensors include temperature sensors, pressure sensors, speed sensors, voltage sensors, current sensors, etc.

[0067] The control unit coordinates the work of all units. It runs a Linux operating system based on a microprocessor (MPU). The control unit manages advanced functions such as task scheduling, error handling, and security policies.

[0068] The data analysis unit has a corresponding algorithm model. Optionally, it is trained using historical data of normal device operation collected by sensors, based on the One-Class SVM algorithm. By adjusting the nu hyperparameter and threshold, a novelty recall rate of 100% was achieved, and the model accuracy was maintained above 50%. The model judges sensor data. If sensor data is judged as novel, meaning that the sensor data has never appeared in the historical data of normal device operation or has no similarity, it indicates that the device is abnormal and the sensor data needs to be reported to the cloud server for alarm and further analysis. If sensor data is judged as normal, it means that the device is operating normally and the sensor data is discarded.

[0069] The data reporting unit connects to cloud services and supports multiple communication methods such as ETH, 5G, and WiFi, as well as multiple communication protocols such as MQTT, HTTP, COAP, and TCP / UDP.

[0070] Of course, a power supply unit can also be included to provide a stable and reliable power supply to all units of the device. This includes various power options such as batteries and power adapters. For mobile or remotely deployed devices, battery capacity and charging efficiency are particularly critical. Therefore, the power supply unit is designed with a battery management system (BMS) in mind. This system monitors battery status, such as voltage, temperature, and remaining charge, to prevent overcharging, over-discharging, and other problems.

[0071] The equipment involved in this application can include, but is not limited to, those in the field of machining, such as CNC machine tools, stamping machines, and injection molding machines, which can ensure machining accuracy and stable equipment operation by monitoring spindle speed, tool wear, pressure parameters, etc.; in chemical production, equipment such as reaction vessels, distillation towers, pumps, and valves can monitor data such as temperature, pressure, liquid level, and flow rate in real time to prevent safety accidents such as leaks and explosions; for power generation equipment, such as boilers and steam turbines in thermal power plants, wind turbines in wind farms, and photovoltaic panels and inverters in solar power plants, data such as power generation efficiency, equipment energy consumption, and operating status can be monitored to improve energy conversion efficiency; in the power transmission and distribution sector, equipment such as transformers, circuit breakers, and switchgear can ensure stable power transmission by monitoring parameters such as voltage, current, and load; and in energy storage equipment, such as lithium battery packs and supercapacitors, battery voltage, temperature, and charge / discharge status can be monitored to optimize the performance and safety of energy storage systems.

[0072] The following describes in detail the device monitoring method based on AI and Industrial Internet of Things (IIoT) that relates to this application. Specific details, such as particular system architectures and technologies, are presented for illustrative purposes and not for limitation, to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application can also be implemented in other embodiments without these specific details.

[0073] It should be understood that, when used in this specification, the term "comprising" indicates the presence of the described feature, integral, step, operation, element, and / or component, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0074] The terms "one embodiment" or "some embodiments" used in this application mean that one or more embodiments of this application include the specific features, structures, or characteristics described in that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this application do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized.

[0075] In embodiments of the present invention, computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include, but are not limited to, object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer. In cases involving a remote computer, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (exemplarily using an Internet service provider for Internet connection).

[0076] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0077] Please see Figure 1 The diagram shows a flowchart of a device monitoring method based on AI and Industrial IoT in a specific embodiment. The method includes:

[0078] S101: Obtain device monitoring operations.

[0079] In some embodiments, monitoring commands input by the user are received through a user interface, or monitoring operations are automatically triggered by time periods or changes in device status. Operation types include specifying the monitoring device range, setting monitoring parameters such as sampling frequency and monitoring duration, and selecting monitoring modes such as real-time monitoring and historical data backtracking. This supports both on-demand monitoring by users and automated monitoring.

[0080] S102: Trigger the device monitoring process based on the device monitoring operation, and obtain the key monitoring dataset based on the device monitoring process; the key monitoring dataset includes: key monitoring data.

[0081] In some embodiments, monitoring operation instructions are transmitted to the monitoring process scheduling module, triggering the start of the device monitoring process. Based on the operation requirements, the process interacts with device sensors, controllers, etc., through industrial IoT communication protocols to acquire device operating parameters, such as temperature, pressure, speed, running / stopping status, fault codes, etc., and integrates them to form a key monitoring dataset.

[0082] In some specific embodiments, step S102 specifically includes:

[0083] Determine the trigger condition type for equipment monitoring operations. Trigger condition types include: periodic triggering, such as automatically generating monitoring operations at preset time intervals; event-driven triggering, such as generating monitoring operations upon receiving equipment abnormal alarm signals, parameter over-limit signals, or external control commands; or manual triggering.

[0084] The system matches the corresponding device monitoring process based on the trigger condition type. The matching process is based on a preset functional mapping relationship between the trigger type and the monitoring process. For example, periodic triggers match regular inspection monitoring processes, event-driven triggers match fault diagnosis monitoring processes, and manual triggers match customized data acquisition monitoring processes.

[0085] Extract sensor data source information related to the current monitoring process. The data source information includes temperature, pressure, speed sensors, Modbus, CAN, Ethernet, and data acquisition frequency requirements, in order to determine the real-time temperature value, pressure fluctuation range, and speed deviation value of the key monitoring data that need to be acquired.

[0086] Based on the data source information, the data acquisition interface is invoked to collect real-time operating parameters during device operation. These parameters are then encapsulated according to a preset data structure to form a key monitoring dataset containing critical monitoring data, ensuring the dataset's integrity and traceability. The data structure can be set to timestamps, parameter values, and device identifiers.

[0087] This embodiment supports three trigger types: periodic, event-driven, and manual, adaptable to diverse monitoring scenarios in industrial environments. By pre-setting the mapping relationship between trigger types and process functions, it achieves automatic matching of monitoring processes, reducing the risk of misoperation and improving system response speed. Sensor data source information is extracted based on process requirements, clearly defining acquisition parameters and frequencies to accurately capture abnormal signals.

[0088] S103: Query the preset key data analysis rule table based on the key monitoring data to obtain the device data parsing parameters of the key monitoring data.

[0089] This embodiment extracts features from key monitoring data, such as data type, timestamp, and device identifier. These extracted features are used as query conditions to retrieve data from a pre-defined key data analysis rule table. The rule table contains parsing rules corresponding to different devices and data types, such as data format conversion rules (binary to decimal, unit conversion, and filtering rules). After querying, the corresponding parsing parameters are obtained. Based on a rule matching mechanism, the applicable rules are quickly located by comparing data features with the index fields of the rule table, and the parameter configurations in the rules are applied to data parsing.

[0090] S104: Query the preset monitoring task sequence based on the key monitoring data to obtain the monitoring process label.

[0091] This embodiment analyzes key monitoring data, including device type and operating status, and queries the corresponding monitoring task process in a preset monitoring task sequence table. The task sequence table defines the priority and analysis process of monitoring tasks under different devices and operating conditions, such as detecting abnormal data first and then analyzing trends. After querying, the monitoring process tag is obtained to identify the current monitoring task type. This embodiment utilizes a task matching and tag mapping mechanism to quickly locate task sequences based on data characteristics, associate task types with process tags, guide the analysis process calls, and optimize the operating efficiency of the monitoring system.

[0092] S105: Call the device running status analysis process based on the monitoring process label.

[0093] This embodiment identifies the monitoring task type based on the tag and calls the corresponding analysis program from the analysis process library, such as a machine learning-based anomaly detection model or a statistical performance analysis model. Different analysis processes are designed for specific tasks; for example, the anomaly detection process is used to identify data anomalies, and the performance evaluation process is used to calculate equipment efficiency indicators. Based on the tag-driven process scheduling mechanism, the appropriate analysis tool is dynamically called through the mapping relationship between tags and analysis processes, achieving precise matching between tasks and algorithms.

[0094] S106: The monitoring equipment is analyzed through the equipment operation status analysis process to obtain the monitoring target data.

[0095] In this embodiment, the invoked analysis process processes key monitoring data, such as identifying abnormal data through machine learning algorithms, predicting equipment performance trends using time series analysis, and determining equipment status (normal / warning / fault) based on rules. After processing, the analysis results are output, such as a list of abnormal data, performance prediction curves, and equipment status conclusions, forming the monitoring target data.

[0096] In some specific embodiments, step S106 specifically includes: aligning the data from real-time sensor data, historical device logs, external environment monitoring data, numerical time-series data, text-based maintenance records, and image-based device appearance photos in the key monitoring data.

[0097] Based on the current operating conditions of the equipment, the anomaly records of the past 30 days, and the last maintenance time, dynamic features strongly correlated with the equipment status are extracted. For example, when the equipment is running at a high load (>80%), the correlation coefficient between temperature data and load is extracted; when the load is low (<30%), the energy proportion of the 100-500Hz frequency band in the vibration spectrum is extracted to ensure that the features are highly adapted to the current operating scenario.

[0098] The system invokes a preset status assessment rule table, which includes three levels of status definitions: normal, warning, and fault. Based on the extracted features, the system performs a graded assessment of the equipment's operating status.

[0099] Based on the current operating conditions, such as load level and ambient temperature, the evaluation threshold is adjusted. For example, the temperature anomaly threshold is reduced from 80℃ to 75℃ under high load. The status level is determined by the correlation between features, such as temperature > 75℃ and vibration energy ratio > 30%, and the status conclusion is output, such as warning - overload risk and temperature anomaly caused by excessive load.

[0100] The status assessment results are correlated with the original key monitoring data, extracted dynamic features, and assessment criteria, such as specific thresholds and related feature combinations, to generate monitoring target data including normal / warning / fault status, temperature peaks, excessive vibration energy percentage, bearing wear, and timestamps, ensuring data traceability and interpretability.

[0101] As can be seen, the perception feature extraction method extracts dynamic features based on the actual operating conditions and historical information of the equipment, avoiding the extraction of irrelevant or inefficient features, ensuring that the extracted features are highly correlated with the current operating status of the equipment, and improving the accuracy of the features in representing the equipment status. Through multi-feature association analysis, the operating status of the equipment can be judged more accurately, reducing the probability of misjudgment and missed judgment, and timely detection of potential equipment problems.

[0102] S107: Analyze the key monitoring data based on the monitoring target data to obtain the monitoring results of the target device.

[0103] This embodiment combines target monitoring data and key monitoring data, and through data correlation analysis, such as linking abnormal data with equipment operation logs, generates trend charts and status dashboards, outputting the monitoring results of the target equipment. The results include information such as equipment operating status description, abnormality cause analysis, and maintenance suggestions, presented in a structured report or visual interface, enabling preventative equipment maintenance.

[0104] In one embodiment of the present invention, based on step S103, the following will provide a possible embodiment and its specific implementation will be described in a non-limiting manner. Step S103 specifically includes:

[0105] Extract multi-dimensional feature tags from key monitoring data. Feature tags include physical quantities such as temperature and pressure, real-time data, historical data, periodic sampling data, equipment number, equipment type, key component identification, and abnormal risk level, forming a structured set of query conditions.

[0106] The preset key data analysis rule table is dynamically filtered based on feature tags. The filtering process combines the current operating conditions of the equipment, such as load status, operating mode, whether there have been any recent abnormalities, and the last maintenance time, to exclude inapplicable rules from the rule table. For example, high-load exclusive parsing rules are not enabled when the load is low, and candidate rules that are strongly related to the current scenario are retained.

[0107] The retained candidate rules are conflict resolved. The conflict resolution is based on the confidence, timeliness and coverage of the rules, and the rule with the highest confidence, the strongest timeliness and the best coverage is selected as the target parsing rule.

[0108] As one implementation of this application, a multi-dimensional rule resolution indicator set is constructed. In addition to the original confidence level, timeliness, and coverage, rule logic is configured, such as whether it is based on a clear physical formula or statistical model and whether the rule supports protocol interface with industrial IoT platforms, SCADA systems, or MES systems, forming an indicator that includes confidence level C, timeliness T, coverage S, interpretability E, and cross-system compatibility I.

[0109] Based on key production equipment, auxiliary monitoring equipment, and current task objectives, dynamic weight values ​​are assigned to each dimension. For example, in a safety early warning task for key production equipment, the confidence level (C) weight is increased to 40%, down from the default of 30%. The interpretability (E) weight is increased to 25%, down from the default of 20%. The cross-system compatibility (I) weight is reduced to 10%, down from the default of 15%. This ensures that the focus of resolution is strongly correlated with the importance of the equipment and the task requirements.

[0110] Based on the weighted index value, index value = C×Wc + T×Wt + S×Ws + E×We + I×Wi, the priority of each candidate rule is calculated; at the same time, the mutual exclusion between rules is detected. For example, if two rules have contradictory parsing logic for the same parameter, such as one rule defining "temperature > 80℃ alarm" and another defining "temperature > 75℃ alarm", the rule with the higher comprehensive score and no mutual exclusion is selected first; if mutually exclusive rules exist, manual review is triggered or historical verification data is called to select the better rule.

[0111] By simulating operating data, such as test sets generated based on historical equipment operating data, the parsing accuracy of the rules in the current scenario is verified. If the verification result deviates from the actual expectation by more than a preset threshold, the coverage of the rule is adjusted or the confidence level is corrected, such as by reducing it by 10%. The corrected rule is re-entered into the candidate pool. If it still fails to meet the standard after multiple verifications, it is downgraded to a backup rule. Finally, the rule that passes the verification and has the highest comprehensive score is selected as the target parsing rule.

[0112] Based on the target parsing rules, the data sampling frequency, feature extraction window length, and unit conversion coefficient defined in the rules are extracted. The parsing parameters are then dynamically corrected according to the actual acquisition accuracy of the key monitoring data. For example, the theoretical sampling frequency is adjusted to the highest frequency that can be acquired in practice. Finally, the device data parsing parameters adapted to the current key monitoring data are obtained.

[0113] As can be seen, by extracting feature labels and dynamically filtering, combined with equipment operating conditions and historical information to exclude inapplicable rules, data parsing errors caused by rule mismatches are avoided, ensuring that the parsing parameters closely match the actual operating status of the equipment. The conflict resolution mechanism comprehensively considers rule confidence, timeliness, and coverage, prioritizing the latest and most suitable rules, enabling the system to adapt to dynamic scenarios such as changes in equipment operating conditions and rule updates, thus improving the generalization ability of the rules. The analysis of features such as anomaly risk levels allows the system to adopt stricter parsing rules for highly sensitive parameters, enabling the processing of critical equipment data and improving the accuracy of fault warning and diagnosis.

[0114] In one embodiment of the present invention, based on step S107, the following will provide a possible embodiment and its specific implementation will be described in a non-limiting manner. Step S107 specifically includes:

[0115] Step S1071: Determine the parsing order of key monitoring data. The parsing order is based on the application scenario of the monitoring target data, such as real-time alarm, trend analysis, fault diagnosis, and the criticality of the data. For example, temperature exceeding the limit that affects equipment safety is placed at the front of the parsing queue, and environmental humidity is placed at the back.

[0116] Step S1072: Match the model tag corresponding to the monitoring target data from the preset model tag library. The matching is based on the key parameter type of the monitoring target data. For example, pressure fluctuation data corresponds to the pressure anomaly detection tag, and speed change data corresponds to the mechanical wear early warning tag, ensuring that the called equipment monitoring model is highly compatible with the current analysis requirements.

[0117] Step S1073: According to the determined parsing order, call the matching device monitoring model to parse the key monitoring data step by step. The parsing process includes feature extraction and preliminary judgment of the data. During the parsing process, the model parameters are dynamically adjusted according to the real-time data characteristics, such as narrowing the threshold range of anomaly detection to improve accuracy.

[0118] Step S1074: Aggregate the equipment features obtained from step-by-step parsing in multiple dimensions according to business requirements. The aggregation method includes time dimension or logical dimension. For example, associate high temperature with excessive speed to determine overload risk, and finally form a target equipment monitoring result covering the whole perspective of equipment operation status.

[0119] In this way, by adjusting the parsing order, the problem of critical data being delayed due to a fixed order, as seen in traditional methods, is avoided, ensuring that highly sensitive parameters related to equipment safety are analyzed first. Precise matching of model labels improves the targeting and accuracy of the parsing; dynamically adjusting model parameters during the parsing process avoids false alarms or missed alarms caused by changes in operating conditions due to fixed thresholds; and it enhances the intelligence level of equipment monitoring and the operability of the results.

[0120] Step S107 also includes the following methods:

[0121] The method for determining the parsing order of key monitoring data is as follows:

[0122] MI ( X , Y )= α ⋅ MI ( X , Y )+ β ⋅ H ( Y | X )+ c ⋅Corr( X , Y )

[0123] in, X To monitor key data feature sets, Y To monitor the target data feature set, H ( Y | X ) is the conditional entropy, Corr( X , Y ) is the Pearson correlation coefficient (a measure of linear correlation). α , β , c corresponding The weighting coefficients are dynamically adjusted based on the monitoring scenario; for example, they are increased in real-time alarm scenarios. c To emphasize linear correlation. This is achieved by calculating the relationship between each candidate data feature and the target feature. MI Improve the values, sort them in descending order to determine the parsing order, and prioritize the parsing of features that are strongly correlated with the target.

[0124] The device monitoring model parses data based on a threshold function, where the threshold function is:

[0125] i ( t )= i 0+ k ⋅exp(−2 s 2( t - t 0)2)

[0126] in, i 0 is the preset base threshold. k The adjustment coefficient is set based on the historical abnormal frequency of the equipment; during high-frequency abnormalities... k >0 to tighten the threshold. t For the current time, t 0 represents a period of high frequency of historical equipment anomalies. s The standard deviation of the time window. When calling the matching model, the anomaly detection threshold of the model is dynamically adjusted. i ( t The model outputs device characteristics, such as an 85% probability of temperature anomalies and a 3.2σ amplitude of sudden speed changes.

[0127] Device features are aggregated using the information entropy formula. The information entropy formula is:

[0128] H ( D )=−∑ i =1 npi ⋅log( pi )+ l ⋅max(0, s 2∑ i =1 m ( xi - m )2− t )

[0129] in, D For the set of device features, pi For the first i The probability of class data appearing l The regularization coefficient is . m For data dimensions, xi For the first i Dimensional data values, m The mean, s 2 represents the variance. t This is the tolerance threshold for abnormal fluctuations. It is calculated... H ( D ) value, if H ( D )> l If an alarm is triggered and a preset entropy threshold is set, the device is determined to be abnormal, and the final monitoring result will be high risk and specific abnormal dimensions; otherwise, it will be normal and a summary of key parameters.

[0130] This embodiment achieves intelligent analysis of key monitoring data and equipment status assessment by integrating information theory, statistics, and a dynamic threshold adjustment mechanism. It uses an improved mutual information formula to calculate the correlation between key monitoring data and target data, comprehensively considering mutual information, conditional entropy, and linear correlation. The weighting coefficients are dynamically adjusted according to the monitoring scenario, and data analysis priority is determined in descending order of correlation to ensure that data strongly correlated with equipment status is processed first.

[0131] The anomaly detection criteria of the equipment monitoring model are dynamically adjusted based on a threshold function θ(t). This function combines historical anomaly patterns of the equipment with time factors, automatically tightening the threshold during periods of high anomaly incidence to improve the sensitivity of anomaly detection. During model analysis, quantitative indicators of equipment characteristics are output.

[0132] Finally, the information entropy formula H(D) is used to aggregate device features. By introducing a regularization term and an anomaly fluctuation tolerance threshold, the uncertainty and volatility of the data are comprehensively evaluated. When the information entropy exceeds a preset threshold, the device is determined to be abnormal, thus outputting target monitoring results covering risk level and specific anomaly dimensions. The entire process is driven by a mathematical model to achieve a closed loop from data correlation analysis to accurate anomaly determination.

[0133] In this way, based on the parsing order determination method, data processing priorities can be adjusted according to factors such as real-time alarms and trend analysis, prioritizing the parsing of data that has the greatest impact on equipment status, thereby improving the processing efficiency and response speed of critical information. The threshold function θ(t) dynamically adjusts the judgment threshold by combining the historical anomaly patterns of the equipment with time changes, automatically increasing detection sensitivity during periods of high anomaly incidence, effectively capturing early fault signs, reducing the risk of missed detections, and improving the accuracy of anomaly detection.

[0134] The information entropy aggregation formula H(D) integrates multi-dimensional features such as data probability distribution and fluctuation level. Through regularization and anomaly tolerance threshold constraints, it achieves a quantitative assessment of equipment status, ensuring the reliability and comprehensiveness of anomaly judgment results. Parameters such as weighting coefficients (α, β, γ), adjustment coefficients (k), and regularization coefficients (λ) can be flexibly adjusted according to different equipment types and operating conditions, enabling the system to adapt to the complex and ever-changing monitoring needs in the Industrial Internet of Things, enhancing its versatility and scalability.

[0135] It should be understood that the sequence number of each step in the above embodiments does not imply 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 invention.

[0136] The following are embodiments of the device monitoring system based on AI and Industrial Internet of Things provided in this disclosure. This system and the device monitoring methods based on AI and Industrial Internet of Things in the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the device monitoring system based on AI and Industrial Internet of Things, please refer to the embodiments of the device monitoring methods based on AI and Industrial Internet of Things described above.

[0137] like Figure 2 As shown, the system includes: a control unit 201, a data acquisition unit 202, a data analysis unit 203, and a data reporting unit 204.

[0138] The data acquisition unit 202 is used to acquire sensor data.

[0139] Control unit 201 is used to coordinate the work of each unit and acquire equipment monitoring operations.

[0140] The data analysis unit 203 is used to trigger the device monitoring process based on the device monitoring operation, and to obtain the key monitoring dataset based on the device monitoring process; the key monitoring dataset includes key monitoring data.

[0141] The system queries the preset key data analysis rule table based on the key monitoring data to obtain the device data parsing parameters of the key monitoring data; it queries the preset monitoring task sequence based on the key monitoring data to obtain the monitoring process label; it calls the device operation status analysis process based on the monitoring process label; it analyzes the monitored device through the device operation status analysis process to obtain the monitoring target data; and it parses the key monitoring data based on the monitoring target data to obtain the monitoring results of the target device.

[0142] The data analysis unit features a One-Class SVM algorithm. When processing key monitoring data in the key monitoring dataset, the data analysis unit can create a pipelined model that includes data normalization and One-Class Support Vector Machine (One-Class SVM). The specific steps are as follows:

[0143] First, StandardScaler is used to standardize the key monitoring data to eliminate the differences in dimensions between different features, so that the data conforms to a normal distribution with a mean of 0 and a variance of 1.

[0144] Then, the OneClassSVM algorithm is called as the core model, and the parameters nu (controlling the proportion of outliers), kernel='rbf' (radial basis function kernel, used to handle nonlinear data) and gamma='auto' (automatically calculate kernel function coefficients) are set.

[0145] The steps to design the `evaluate_model` function for training a model and evaluating its performance are as follows:

[0146] The input parameters are the pipeline model and the model name.

[0147] Use pipeline.fit(X_train) to fit the key monitoring dataset, allowing the model to learn the distribution characteristics of normal key monitoring data;

[0148] The distance score between the key data of test monitoring and the distribution of key data of normal monitoring is calculated by pipeline.decision_function(X_test). The higher the score, the closer the data is to the normal distribution.

[0149] Based on the preset threshold THRESHOLD, the test set of key monitoring data is predicted: if the score is greater than the threshold, it is judged as normal data and labeled; otherwise, it is judged as abnormal data and labeled -1.

[0150] The evaluation indicators include:

[0151] Recall: Measures a model's ability to correctly identify outlier data;

[0152] Precision: Measures the proportion of data that the model classifies as anomalous that are actually anomalous.

[0153] F1 score: A balanced metric that combines recall and precision.

[0154] Call the `evaluate_model` function, passing in the constructed pipeline model `pipeline_with_scaler` and the model name "using StandardScaler", and output the evaluation results, including recall, precision, and F1 score, to determine the model's performance in detecting outliers.

[0155] The data reporting unit 204 is connected to the cloud server 205 and reports the monitoring results of the target device to the cloud server through the network.

[0156] like Figure 3 As shown, this application also provides an electronic device, including a display module 103, a memory 102, a processor 101, and a computer program stored in the memory and executable on the processor 101. When the processor 101 executes the program, it implements the steps of a device monitoring method based on AI and industrial IoT.

[0157] In embodiments of the present invention, electronic devices include, but are not limited to, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments described and / or claimed herein.

[0158] In this embodiment, processor 101 may be implemented using at least one of an application-specific integrated circuit, a programmable logic device, a field-programmable gate array, a processor, a controller, a microcontroller, a microprocessor, or an electronic unit designed to perform the functions described herein. In some cases, such an implementation may be implemented within a controller. For software implementation, implementations such as processes or functions may be implemented with separate software modules that allow the performance of at least one function or operation. Software code may be implemented by a software application (or program) written in any suitable programming language, and the software code may be stored in memory and executed by the controller.

[0159] The display module 103 is used to display information input by the user or information provided to the user. The display module 103 may include a display panel, which may be configured in the form of a liquid crystal display, an organic light-emitting diode, or the like.

[0160] The memory 102 can be used to store software programs and various data. The memory 102 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0161] This application also provides a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the AI- and Industrial Internet of Things-based device monitoring method.

[0162] The storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example,, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0163] In a storage medium, a readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying readable program code. This propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0164] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A device monitoring method based on AI and Industrial Internet of Things, characterized in that, The methods include: S101: Obtain device monitoring operations; S102: Trigger the device monitoring process based on the device monitoring operation, and obtain the key monitoring dataset based on the device monitoring process; The monitoring of key datasets includes monitoring key data; S103: Query the preset key data analysis rule table based on the key monitoring data to obtain the device data parsing parameters of the key monitoring data; Extract feature tags from key monitoring data. The feature tags include data type, time attribute, device association attribute and anomaly risk level, forming a set of query conditions. Filter the preset key data analysis rule table based on feature tags; The retained candidate rules are conflict resolved. The conflict resolution is based on the confidence, timeliness and coverage of the rules, and the rule with the highest confidence, the strongest timeliness and the best coverage is selected as the target parsing rule. Based on the target parsing rules, the parsing parameters defined in the rules are extracted, and the parsing parameters are corrected according to the actual acquisition accuracy of the key monitoring data, so as to finally obtain the device data parsing parameters that are adapted to the current key monitoring data. S104: Query the preset monitoring task sequence based on key monitoring data to obtain monitoring process tags; S105: Invoke the device running status analysis process based on the monitoring process label; S106: Analyze the monitored equipment through the equipment operation status analysis process to obtain the monitoring target data; Data fusion preprocessing is used to align and integrate data from different channels and formats. Based on the current operating condition and historical context information of the equipment, dynamic features closely related to the equipment status are extracted from the fused dataset, making the features adaptable to different scenarios. Using a pre-set status assessment rule table and an adaptive assessment model, combined with extracted dynamic features, the assessment threshold is dynamically adjusted according to real-time operating conditions. The equipment operating status is graded and assessed by integrating the correlation of multiple features, and status conclusions and key influencing factors are obtained. The status assessment results are correlated with the original data, features, and assessment criteria to generate structured monitoring target data; S107: Analyze the key monitoring data based on the monitoring target data to obtain the monitoring results of the target device.

2. The equipment monitoring method based on AI and Industrial Internet of Things according to claim 1, characterized in that, Step S102 specifically includes: Determine the trigger condition type for device monitoring operations. Trigger condition types include: periodic triggering, event-driven triggering, or manual active triggering. The device monitoring process is matched according to the trigger condition type, and the matching process is based on the preset functional mapping relationship between the trigger type and the monitoring process; Extract sensor data source information related to the current monitoring process; Based on the data source information, the data acquisition interface is called to collect real-time operating parameters during the device operation process, and the data is encapsulated according to the preset data structure to form a key monitoring dataset containing key monitoring data.

3. The equipment monitoring method based on AI and Industrial Internet of Things according to claim 1, characterized in that, Step S107 specifically includes: Determine the parsing order of key monitoring data. The parsing order is determined based on the application scenario and the criticality of the monitoring target data. Match the model tags corresponding to the monitored target data from the preset model tag library; According to the established parsing order, the matching device monitoring model is invoked to parse the key monitoring data step by step. The parsing process includes feature extraction and preliminary judgment of the data, and the model parameters are adjusted according to the real-time data characteristics during the parsing process. The device features obtained through step-by-step parsing are aggregated according to business requirements. The aggregation method includes time and logical dimensions, ultimately forming a target device monitoring result that covers the device's operating status.

4. The equipment monitoring method based on AI and Industrial Internet of Things according to claim 3, characterized in that, Step S107 specifically includes: The method for determining the parsing order of key monitoring data is as follows: MI ( X , Y )= α ⋅ MI ( X , Y )+ β ⋅ H ( Y ∣ X )+ γ ⋅Corr( X , Y ) in, X To monitor key data feature sets, Y To monitor the target data feature set, H ( Y | X ) is the conditional entropy, Corr( X , Y ) represents the Pearson correlation coefficient. α , β , γ corresponds These are the weighting coefficients; The device monitoring model parses data based on a threshold function, where the threshold function is: θ ( t )= θ 0+ k ⋅exp(−2 σ 2( t − t 0)2) in, θ 0 is the preset base threshold. k For adjustment coefficients, t For the current time, t 0 represents a period of high frequency of historical equipment anomalies. σ The standard deviation of the time window; Device features are aggregated using the information entropy formula, which is: H ( D )=−∑ i =1 npi ⋅log( pi )+ λ ⋅max(0, σ 2∑ i =1 m ( xi − μ )2− τ ) in, D For the set of device features, pi For the first i The probability of class data appearing λ The regularization coefficient is . m For data dimensions, xi For the first i Dimensional data values, μ The mean, σ 2 represents the variance. τ This is the tolerance threshold for abnormal fluctuations.

5. The equipment monitoring method based on AI and Industrial Internet of Things according to claim 1, characterized in that, The specific steps for conflict resolution of the retained candidate rules include: Construct a multi-dimensional rule resolution indicator set, forming an indicator set that includes confidence level C, timeliness T, coverage S, interpretability E, and cross-system compatibility I: Dynamic weight values ​​are assigned to each dimension based on device type and current task objective; Based on the weighted index value, index value = C×Wc+T×Wt+S×Ws+E×We+I×Wi, calculate the priority of each candidate rule; at the same time, prioritize the task target with the higher comprehensive score. The accuracy of rule parsing in the current scenario is verified by simulating working condition data. If the verification result deviates from the actual expectation by more than a preset threshold, the coverage of the rule will be adjusted and the candidate will re-enter the candidate pool. If the rule fails to meet the standard after multiple verifications, it will be downgraded to a backup rule. Finally, the rule that passes verification and has the highest overall score will be selected as the target parsing rule.

6. A device monitoring system based on AI and Industrial Internet of Things, characterized in that, The system is used to implement the device monitoring method based on AI and industrial IoT as described in any one of claims 1 to 5; The system includes: a control unit, a data acquisition unit, a data analysis unit, and a data reporting unit; The data acquisition unit is used to acquire sensor data; The control unit is used to coordinate the work of each unit and acquire equipment monitoring data. The data analysis unit is used to trigger the device monitoring process based on device monitoring operations, and to obtain the key monitoring dataset based on the device monitoring process; the key monitoring dataset includes key monitoring data. The system queries the preset key data analysis rule table based on the key monitoring data to obtain the device data parsing parameters of the key monitoring data; it queries the preset monitoring task sequence based on the key monitoring data to obtain the monitoring process label; it calls the device operation status analysis process based on the monitoring process label; it analyzes the monitored device through the device operation status analysis process to obtain the monitoring target data; and it parses the key monitoring data based on the monitoring target data to obtain the monitoring results of the target device. The data reporting unit communicates with the cloud server and reports the monitoring results of the target device to the cloud server via the network.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the device monitoring method based on AI and Industrial Internet of Things as described in any one of claims 1 to 5.

8. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the device monitoring method based on AI and Industrial Internet of Things as described in any one of claims 1 to 5.

Images