Micro-monitoring alarm method and system based on multi-mode identification

Abstract

The invention discloses a micro-monitoring alarm method and system based on multi-mode identification. The method comprises the following steps: acquiring monitoring data; analyzing according to the monitoring data to obtain a feature vector; according to feature vector analysis, obtaining an anomaly degree, an anomaly influence range, an anomaly duration and an anomaly occurrence frequency; according to the abnormality degree, the abnormality influence range, the abnormality duration and the abnormality occurrence frequency evaluation, obtaining an abnormality risk value; adjusting according to the exception duration of different equipment types to obtain an exception judgment threshold value; judging according to the abnormal risk value and an abnormal judgment threshold value; if the abnormal risk value is greater than the abnormal judgment threshold value, generating a preliminary alarm signal; according to the preliminary alarm signal, the abnormal influence range and the equipment importance degree arrangement, obtaining a graded alarm information queue; and analyzing according to the graded alarm information queue, the importance degree of the equipment and the abnormal duration to obtain an equipment abnormal alarm signal. According to the method, the safety of equipment monitoring can be improved through a multi-stage alarm processing mechanism.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring and security for industrial equipment, and particularly to a micro-monitoring alarm method and system based on multi-modal recognition. Background Art

[0002] In a complex industrial production environment, the operating states of various devices and systems need to be monitored and analyzed in real time to ensure the safety and stability of the production process. However, due to the complexity and variability of the production environment, traditional single-modal monitoring methods are difficult to comprehensively and accurately reflect the actual operating conditions of devices and systems, and are prone to false alarms and missed alarms.

[0003] To address this issue, there is an urgent need to study an intelligent monitoring and alarm technology based on multi-modal fusion. This technology needs to comprehensively utilize different modal data collected by multiple sensors, such as vibration, temperature, pressure, sound, etc., and through data fusion and intelligent analysis algorithms, achieve a comprehensive perception and accurate judgment of the operating states of devices and systems. At the same time, reasonable data collection frequencies and time windows need to be designed for different monitoring objects and working conditions to balance real-time performance and computational efficiency. Another key issue is how to achieve intelligent alarm decision-making and hierarchical processing based on the results of multi-modal data analysis. Traditional alarm triggering conditions are usually based on the threshold judgment of a single indicator, which is difficult to adapt to complex and changing industrial scenarios. Therefore, it is necessary to study alarm decision-making algorithms with strong adaptability and good interpretability, which can dynamically adjust alarm triggering conditions according to the comprehensive analysis results of multi-modal data, and perform hierarchical processing and priority sorting of alarm events according to factors such as the degree of abnormality and the scope of influence, so as to improve the accuracy and effectiveness of alarms, reduce the occurrence of false alarms and missed alarms, and enhance the safety and reliability of industrial production.

[0004] In an existing technology, a single-modal monitoring method is adopted. For example, only relying on vibration sensors or temperature sensors to monitor the operating state of devices. The implementation process often includes installing corresponding sensors on the devices, collecting data and analyzing it by setting fixed thresholds. When the data of the sensors exceeds the set thresholds, the system will trigger an alarm.

[0005] However, in the existing technology, due to relying only on single-modal data, which is limited by its own characteristics and unable to capture multi-dimensional information, especially in a complex and dynamic industrial environment, the threshold judgment of a single indicator often cannot accurately reflect the actual health state of the device, thus affecting the safety and reliability of production. Summary of the Invention

[0006] The present invention provides a micro-monitoring alarm method and system based on multi-modal recognition to achieve priority sorting of abnormal events through a multi-level alarm processing mechanism, and improve the safety and reliability of device monitoring.

[0007] In a first aspect, to solve the above technical problems, the present invention provides a micro-monitoring and alarm method based on multi-modal recognition, including:

[0008] Obtain real-time monitoring data;

[0009] Perform feature analysis on the real-time monitoring data to obtain a state feature vector;

[0010] Perform anomaly analysis on the state feature vector to obtain the anomaly degree, the anomaly influence range, the anomaly duration, and the anomaly occurrence frequency;

[0011] Evaluate according to the anomaly degree, the anomaly influence range, the anomaly duration, and the anomaly occurrence frequency to obtain an anomaly risk value;

[0012] Adjust according to the anomaly duration of different device types to obtain an anomaly judgment threshold;

[0013] Judge according to the anomaly risk value and the anomaly judgment threshold; if the anomaly risk value is greater than the anomaly judgment threshold, generate a preliminary alarm signal; wherein, the preliminary alarm signal includes an alarm level, an alarm reason, and an alarm time;

[0014] Arrange in priority according to the alarm level, the anomaly influence range, and the pre-stored device importance level to obtain a hierarchical alarm information queue;

[0015] Perform decision-making analysis according to the hierarchical alarm information queue, the pre-stored device importance level, and the anomaly duration to obtain a device anomaly alarm signal.

[0016] In a realizable manner of the first aspect, the real-time monitoring data refers to data obtained by monitoring according to different acquisition strategies based on the evaluation result of the health status of the device.

[0017] In a realizable manner of the first aspect, the performing anomaly analysis on the state feature vector to obtain the anomaly degree, the anomaly influence range, the anomaly duration, and the anomaly occurrence frequency includes:

[0018] Analyze the state feature vector by using a knowledge-based intelligent analysis algorithm to obtain an anomaly evaluation function;

[0019] Calculate the membership degree according to the anomaly evaluation function to obtain the anomaly degree;

[0020] Judge according to the abnormal degree and a preset abnormal degree threshold; if the abnormal degree is greater than the preset abnormal degree threshold, it is determined that the device is abnormal, and the influence range, duration and frequency of the abnormal occurrence are counted to obtain the abnormal influence range, abnormal duration and abnormal occurrence frequency; if the abnormal degree is less than the preset abnormal degree threshold, it is determined that the device is normal.

[0021] In one implementable manner of the first aspect, the analysis of the state feature vector by using a knowledge-based intelligent analysis algorithm to obtain an abnormal evaluation function includes:

[0022] The abnormal evaluation function is calculated by the following formula:

[0023]

[0024] where, f i (x i ) represents the abnormal evaluation function of the i-th state feature vector, x i is the current value of the i-th state feature vector, μ i is the normal mean value of the i-th state feature vector, and σ i is the standard deviation of the i-th state feature vector.

[0025] In one implementable manner of the first aspect, the evaluation of the abnormal degree, the abnormal influence range, the abnormal duration and the abnormal occurrence frequency to obtain an abnormal risk value includes:

[0026] The abnormal risk value is calculated by the following formula:

[0027] R = w 1 A + w 2 S + w 3 D + w 4 F

[0028] where, R represents the abnormal risk value, A represents the abnormal degree, S represents the abnormal influence range, D represents the abnormal duration, F represents the occurrence frequency, and w 1 is the abnormal degree weight coefficient, w 2 is the abnormal influence range weight coefficient, w 3 is the abnormal duration weight coefficient, w 4 is the occurrence frequency weight coefficient.

[0029] In one implementable manner of the first aspect, it is characterized in that the prioritization according to the alarm level, the abnormal influence range and the pre-stored device importance degree to obtain a hierarchical alarm information queue includes:

[0030] Calculate the score according to the alarm level, the abnormal influence range, and the importance level of the device stored in advance, to obtain the priority score of the alarm event;

[0031] Sort in descending order according to the priority score of the alarm event to obtain a sequence of alarm events sorted by priority;

[0032] Judge the priority score of the alarm event in the sequence of alarm events sorted by priority; if the priority score of the alarm event is greater than the preset score threshold, determine that the alarm event is of high priority; otherwise it is of low priority;

[0033] Sort according to the high priority and the low priority of the alarm event to obtain a queue of classified alarm information.

[0034] In one implementable manner of the first aspect, the calculating the priority score of the alarm event according to the alarm level, the abnormal influence range, and the importance level of the device stored in advance includes:

[0035] Calculate the priority score of the alarm event through the following formula:

[0036] P = w′ 1 L + w 2 S + w′ 3 I

[0037] where P represents the priority score of the alarm event, L represents the alarm level, S represents the abnormal influence range, I represents the importance level of the device stored in advance, w' 1 is the weight coefficient of the alarm level, w 2 is the weight coefficient of the abnormal influence range, w' 3 is the weight coefficient of the importance level of the device stored in advance.

[0038] In one implementable manner of the first aspect, the performing decision analysis according to the queue of classified alarm information, the importance level of the device stored in advance, and the abnormal duration to obtain a device abnormal alarm signal includes:

[0039] Perform multi-level factor analysis on the queue of classified alarm information by using a preset alarm decision algorithm to obtain an alarm information sending method;

[0040] Select a receiving object according to the importance level of the device stored in advance to obtain a receiving object of the alarm information;

[0041] Set according to the abnormal duration to obtain the sending frequency of the alarm information;

[0042] Combining according to the alarm information sending method, the alarm information receiving object, and the alarm information sending frequency to obtain an equipment abnormal alarm signal.

[0043] In a second aspect, the present invention provides a micro-monitoring alarm system based on multi-modal recognition, including:

[0044] A data acquisition module for acquiring real-time monitoring data;

[0045] A feature analysis module for performing feature analysis on the real-time monitoring data to obtain a state feature vector;

[0046] An abnormality analysis module for performing abnormality analysis on the state feature vector to obtain the degree of abnormality, the scope of abnormal influence, the duration of abnormality, and the frequency of abnormality occurrence;

[0047] A risk assessment module for performing an assessment based on the degree of abnormality, the scope of abnormal influence, the duration of abnormality, and the frequency of abnormality occurrence to obtain an abnormal risk value;

[0048] A threshold determination module for adjusting according to the duration of abnormality of different device types to obtain an abnormality judgment threshold;

[0049] An abnormality judgment module for making a judgment based on the abnormal risk level and the abnormality judgment threshold; if the abnormal risk value is greater than the abnormality judgment threshold, a preliminary alarm signal is generated; wherein, the preliminary alarm signal includes an alarm level, an alarm reason, and an alarm time;

[0050] An alarm sorting module for preferentially arranging according to the alarm level, the scope of abnormal influence, and the importance of the device stored in advance to obtain a hierarchical alarm information queue;

[0051] A signal generation module for performing decision analysis based on the hierarchical alarm information queue, the importance of the device stored in advance, and the duration of abnormality to obtain an equipment abnormal alarm signal.

[0052] In an implementable manner of the second aspect, the real-time monitoring data refers to data obtained by monitoring according to different acquisition strategies based on the health status evaluation result of the device.

[0053] In an implementable manner of the second aspect, the performing abnormality analysis on the state feature vector to obtain the degree of abnormality, the scope of abnormal influence, the duration of abnormality, and the frequency of abnormality occurrence includes:

[0054] Analyzing the state feature vector using a knowledge-based intelligent analysis algorithm to obtain an abnormality evaluation function;

[0055] Calculate the membership degree according to the abnormal evaluation function to obtain the degree of abnormality;

[0056] Make a judgment based on the degree of abnormality and a preset threshold of the degree of abnormality; if the degree of abnormality is greater than the preset threshold of the degree of abnormality, it is determined that the device is abnormal, and the influence range, duration, and frequency of the occurrence of the abnormality are counted to obtain the abnormal influence range, abnormal duration, and abnormal occurrence frequency; if the degree of abnormality is less than the preset threshold of the degree of abnormality, it is determined that the device is normal.

[0057] In one implementation manner of the second aspect, the analysis using the knowledge-based intelligent analysis algorithm according to the state feature vector to obtain the abnormal evaluation function includes:

[0058] The abnormal evaluation function is calculated by the following formula:

[0059]

[0060] where, f i (x i ) represents the abnormal evaluation function of the i-th state feature vector, x i is the current value of the i-th state feature vector, μ i is the normal mean value of the i-th state feature vector, and σ i is the standard deviation of the i-th state feature vector.

[0061] In one implementation manner of the second aspect, the evaluation according to the degree of abnormality, the abnormal influence range, the abnormal duration, and the abnormal occurrence frequency to obtain the abnormal risk value includes:

[0062] The abnormal risk value is calculated by the following formula:

[0063] R = w 1 A + w 2 S + w 3 D + w 4 F

[0064] where, R represents the abnormal risk value, A represents the degree of abnormality, S represents the abnormal influence range, D represents the abnormal duration, F represents the occurrence frequency, and w 1 is the weight coefficient of the degree of abnormality, w 2 is the weight coefficient of the abnormal influence range, w 3 is the weight coefficient of the abnormal duration, and w 4 is the weight coefficient of the occurrence frequency.

[0065] In an implementable manner of the second aspect, it is characterized in that the prioritization is performed according to the alarm level, the abnormal influence range, and the pre-stored importance level of the device to obtain a hierarchical alarm information queue, including:

[0066] Calculate a priority score for the alarm event according to the alarm level, the abnormal influence range, and the pre-stored importance level of the device;

[0067] Perform a descending order sorting according to the priority score of the alarm event to obtain a prioritized alarm event sequence;

[0068] Judge the priority score of the alarm event in the prioritized alarm event sequence; if the priority score of the alarm event is greater than a preset score threshold, determine that the alarm event is of high priority; otherwise, it is of low priority;

[0069] Sort according to the high priority and the low priority of the alarm event to obtain a hierarchical alarm information queue.

[0070] In an implementable manner of the second aspect, the calculation of the priority score of the alarm event according to the alarm level, the abnormal influence range, and the pre-stored importance level of the device includes:

[0071] The priority score of the alarm event is calculated through the following formula:

[0072] P = w′ 1 L + w 2 S + w′ 3 I

[0073] where P represents the priority score of the alarm event, L represents the alarm level, S represents the abnormal influence range, I represents the pre-stored importance level of the device, w' 1 is the alarm level weight coefficient, w 2 is the abnormal influence range weight coefficient, w' 3 is the pre-stored importance level weight coefficient of the device.

[0074] In an implementable manner of the second aspect, the decision analysis according to the hierarchical alarm information queue, the pre-stored importance level of the device, and the abnormal duration to obtain a device abnormal alarm signal includes:

[0075] Perform multi-factor analysis according to the hierarchical alarm information queue using a preset alarm decision algorithm to obtain an alarm information sending method;

[0076] Select a recipient for the alarm information according to the pre-stored importance level of the device to obtain a recipient of the alarm information;

[0077] Set according to the abnormal duration to obtain the alarm message sending frequency;

[0078] Combine according to the alarm message sending method, the alarm message receiving object, and the alarm message sending frequency to obtain an equipment abnormal alarm signal.

[0079] In a third aspect, the present invention further provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the method for micro-monitoring alarm based on multi-modal recognition described in any one of the above is implemented.

[0080] In a fourth aspect, the present invention further provides a computer-readable storage medium. The computer-readable storage medium includes a stored computer program. When the computer program runs, the device where the computer-readable storage medium is located is controlled to execute the method for micro-monitoring alarm based on multi-modal recognition described in any one of the above.

[0081] Compared with the prior art, the present invention has the following beneficial effects:

[0082] The present invention discloses a method for micro-monitoring alarm based on multi-modal recognition, including obtaining real-time monitoring data; performing feature analysis on the real-time monitoring data to obtain a state feature vector; performing abnormal analysis on the state feature vector to obtain the abnormal degree, the abnormal influence range, the abnormal duration, and the abnormal occurrence frequency; performing an evaluation based on the abnormal degree, the abnormal influence range, the abnormal duration, and the abnormal occurrence frequency to obtain an abnormal risk value; adjusting according to the abnormal duration of different device types to obtain an abnormal judgment threshold; judging based on the abnormal risk value and the abnormal judgment threshold; if the abnormal risk value is greater than the abnormal judgment threshold, generating a preliminary alarm signal; where the preliminary alarm signal includes an alarm level, an alarm reason, and an alarm time; performing a priority arrangement based on the alarm level, the abnormal influence range, and the pre-stored equipment importance level to obtain a hierarchical alarm information queue; performing a decision analysis based on the hierarchical alarm information queue, the pre-stored equipment importance level, and the abnormal duration to obtain an equipment abnormal alarm signal.

[0083] The present invention collects multi-source heterogeneous data such as vibration, temperature, pressure, and sound of a device, and performs real-time fusion analysis to generate a feature vector reflecting the comprehensive operating state of the device. Combining a knowledge base and intelligent algorithms, the present invention can accurately judge the degree of abnormality and potential risks of the device, and dynamically adjust the alarm threshold according to factors such as device type and abnormal duration. Through a multi-level alarm processing mechanism, the present invention can prioritize abnormal events and intelligently select alarm methods and recipients according to the alarm level, influence range, etc. The present invention significantly improves the real-time performance and accuracy of device monitoring, realizes early warning and precise alarm of abnormalities, effectively reduces the risk of device failures, and improves the safety and efficiency of industrial production. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 is a schematic flowchart of a micro-monitoring alarm method based on multi-modal recognition provided by the first embodiment of the present invention;

[0085] Figure 2 is a schematic structural diagram of a micro-monitoring alarm system based on multi-modal recognition provided by the second embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0086] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0087] Referring to Figure 1 , the first embodiment of the present invention provides a micro-monitoring alarm method based on multi-modal recognition, including the following steps:

[0088] S1, Obtain real-time monitoring data;

[0089] S2, Perform feature analysis based on the real-time monitoring data to obtain a state feature vector;

[0090] S3, Perform abnormality analysis based on the state feature vector to obtain the degree of abnormality, the scope of abnormal influence, the duration of abnormality, and the frequency of abnormal occurrence;

[0091] S4, Evaluate based on the degree of abnormality, the scope of abnormal influence, the duration of abnormality, and the frequency of abnormal occurrence to obtain an abnormal risk value;

[0092] S5, Adjust according to the duration of abnormality of different device types to obtain an abnormality judgment threshold;

[0093] S6. Make a judgment based on the abnormal risk value and the abnormal judgment threshold; if the abnormal risk value is greater than the abnormal judgment threshold, generate a preliminary alarm signal; wherein, the preliminary alarm signal includes an alarm level, an alarm reason, and an alarm time;

[0094] S7. Perform a priority arrangement according to the alarm level, the abnormal influence range, and the pre-stored importance level of the equipment to obtain a hierarchical alarm information queue;

[0095] S8. Perform a decision-making analysis according to the hierarchical alarm information queue, the pre-stored importance level of the equipment, and the abnormal duration to obtain an equipment abnormal alarm signal.

[0096] In step S1, obtain real-time monitoring data.

[0097] Preferably, obtain real-time monitoring data collected by sensors such as vibration, temperature, pressure, and sound. For different equipment types and working conditions, set corresponding data collection time windows and frequencies, taking into account computational efficiency while ensuring real-time performance. Obtain real-time monitoring data collected by sensors such as vibration, temperature, pressure, and sound, and dynamically adjust the data collection time window and frequency parameters for different equipment types and working conditions. By analyzing historical monitoring data, use machine learning algorithms to establish a mapping model between equipment types, working conditions, and optimal collection parameters to achieve adaptive optimization of the data collection strategy. Dynamically adjust the data collection time window and frequency according to changes in equipment types and working conditions, reducing the frequency of data collection and transmission while meeting real-time requirements, thereby reducing energy consumption and calculations. Preprocess the collected monitoring data, including denoising, outlier filtering, etc., to improve data quality and reduce subsequent computational workload. Use incremental learning algorithms to continuously optimize the mapping model between equipment types, working conditions, and collection parameters based on newly collected monitoring data, improving the accuracy and adaptability of the model. Set different data compression algorithms and compression ratios for different types of monitoring data, such as vibration, temperature, pressure, and sound, reducing data transmission and storage space while ensuring information integrity. Based on the equipment health status assessment results, dynamically adjust the monitoring data collection strategy, appropriately reducing the collection frequency for equipment in good health, and increasing the collection frequency for equipment with high failure risks to achieve reasonable allocation of monitoring resources.

[0098] Exemplarily, in industrial equipment monitoring, dynamically adjusting the data acquisition strategy is crucial for improving monitoring efficiency and reducing costs. Taking a large manufacturing enterprise as an example, there are various types of equipment on its production line, such as lathes, milling machines, and grinders. For different equipment types and operating conditions, adopting an adaptive data acquisition method can significantly improve the monitoring effect. For lathes, a longer acquisition time window and a lower acquisition frequency can be used during normal operation, such as collecting vibration data every 10 minutes. However, when abnormal vibration is detected, the system will automatically shorten the acquisition window to 1 minute and increase the acquisition frequency to once per second to capture fault characteristics in a timely manner. By analyzing historical data, a mapping model between equipment types, operating conditions, and optimal acquisition parameters can be established. For example, for milling machines, it is found that the acquisition frequencies of temperature and vibration data need to be increased during high-speed cutting, while they can be appropriately reduced during low-speed machining. This mapping relationship can be established through machine learning algorithms such as decision trees or random forests to achieve intelligent adjustment of the acquisition strategy. Data preprocessing is a key step in improving data quality. For the sound data collected by grinders, wavelet transform can be used to remove background noise, and a moving average filter can be used to smooth the data curve. This can not only improve the accuracy of subsequent analysis but also reduce the amount of data stored and transmitted. The application of incremental learning algorithms enables the model to be continuously optimized.

[0099] In step S2, feature analysis is performed based on the real-time monitoring data to obtain a state feature vector.

[0100] In a specific embodiment, multi-source heterogeneous monitoring data is input into a data fusion module, and a preset multi-modal data fusion algorithm is used to extract key features of data such as vibration, temperature, pressure, and sound, and perform correlation analysis at the feature level to generate a state feature vector that comprehensively reflects the operating state of the equipment. Obtain multi-source heterogeneous monitoring data, including vibration data, temperature data, pressure data, and sound data. For each type of monitoring data, use the corresponding feature extraction algorithm to extract key features that can reflect the operating state of the equipment. Normalize the extracted vibration features, temperature features, pressure features, and sound features to unify feature data of different scales and dimensions to the same scale and dimension, facilitating subsequent feature fusion and correlation analysis. Use a multi-modal data fusion algorithm to fuse the normalized vibration features, temperature features, pressure features, and sound features to obtain multi-modal fusion features that comprehensively reflect the operating state of the equipment.

[0101] In step S3, anomaly analysis is performed based on the state feature vector to obtain the anomaly degree, anomaly influence range, anomaly duration, and anomaly occurrence frequency.

[0102] In the above step S3, for performing anomaly analysis based on the state feature vector to obtain the anomaly degree, anomaly influence range, anomaly duration, and anomaly occurrence frequency, the following steps are specifically included:

[0103] S31. Analyze according to the state feature vector by using a knowledge-based intelligent analysis algorithm to obtain an anomaly evaluation function.

[0104] It should be noted that the anomaly evaluation function is calculated through the following formula:

[0105]

[0106] where f i (x i ) represents the anomaly evaluation function of the i-th state feature vector, x i is the current value of the i-th state feature vector, μ i is the normal mean of the i-th state feature vector, and σ i is the standard deviation of the i-th state feature vector.

[0107] In this embodiment, this function is based on the Gaussian distribution and is used to evaluate the degree to which the eigenvalue deviates from the normal range. If the anomaly degree value exceeds the preset threshold, it is determined that the device has an anomaly, and the influence range and duration of the anomaly occurrence are further analyzed.

[0108] S32. Calculate the membership degree according to the anomaly evaluation function to obtain the anomaly degree;

[0109] S33. Make a judgment according to the anomaly degree and the preset anomaly degree threshold; if the anomaly degree is greater than the preset anomaly degree threshold, it is determined that the device has an anomaly, and the influence range, duration, and frequency of the anomaly occurrence are counted to obtain the anomaly influence range, anomaly duration, and anomaly occurrence frequency; if the anomaly degree is less than the preset anomaly degree threshold, it is determined that the device is normal.

[0110] It should be noted that in the above steps S31 to S33, the specific implementation process includes: according to the generated comprehensive state feature vector of the device, using a knowledge-based intelligent analysis algorithm, calculating the anomaly degree under the current operating condition of the device, judging the influence range and duration of the anomaly occurrence, and combining the anomaly occurrence frequency to obtain the risk level of potential faults or abnormal operations of the device. According to the device operation data, extract the multi-dimensional feature vectors reflecting the comprehensive state of the device to form a device state feature set. Use a knowledge-based intelligent analysis algorithm to analyze the device state feature set and calculate the anomaly degree value of the device under the current operating condition.

[0111] Exemplarily, the multi-dimensional feature vector extraction can comprehensively reflect the operating condition of the equipment. Taking a large industrial pump as an example, features can be extracted from aspects such as vibration, temperature, pressure, and flow rate. Vibration features may include root mean square value, peak factor, etc.; temperature features may include temperature values and their change rates at key parts; pressure features may include differential pressure between inlet and outlet, pressure fluctuation amplitude, etc.; flow rate features may include flow rate stability, efficiency, etc. These features together constitute a multi-dimensional feature vector reflecting the pump operating state. The knowledge-based intelligent analysis algorithm can effectively process these multi-dimensional features. For example, a fuzzy inference system can be adopted to encode expert experience into a fuzzy rule base. By comparing the extracted feature values with the normal operating range, the membership degree is calculated, and then the degree of abnormality is inferred. Suppose the normal root mean square value range of vibration is 0 - 5 mm / s, and the current value is 7 mm / s, then the abnormal membership degree is 0.6. Combining the membership degrees of other features, the comprehensive degree of abnormality is finally obtained, such as 0.75. If the preset threshold is 0.7, it is determined that the equipment is abnormal. After determining the abnormality, it is necessary to analyze its influence range and duration. The influence range can be determined by tracking the propagation of abnormal features among various components of the equipment. For example, bearing abnormality will cause an increase in shaft vibration, which in turn affects the pump body and pipeline. The duration can be determined by continuously monitoring the change of the degree of abnormality. If the abnormality persists for more than the preset time, such as 30 minutes, maintenance intervention is required. The statistical analysis of the abnormality occurrence frequency helps to evaluate the overall health condition of the equipment. By analyzing historical data, statistical results such as "3 minor abnormalities and 1 serious abnormality occur in every 100 hours of operation" can be obtained. This frequency analysis can reveal the equipment degradation trend and provide a basis for predictive maintenance. The risk assessment model comprehensively considers factors such as the degree of abnormality, influence range, duration, and occurrence frequency, and calculates the risk level of the equipment. The weighted summation method can be adopted to assign weights to each factor. For example, the weight of the degree of abnormality is 0.4, the influence range is 0.3, the duration is 0.2, and the occurrence frequency is 0.1. Suppose the standardized scores of each factor are 0.75, 0.6, 0.8, and 0.5 respectively, then the comprehensive risk score is 0.69. The risk level can be divided into three levels: low, medium, and high, corresponding to the score ranges of 0 - 0.3, 0.3 - 0.7, and 0.7 - 1. In this example, the risk level is medium, indicating that the equipment status needs to be closely monitored and corresponding maintenance plans need to be formulated. This multi-dimensional equipment status monitoring and risk assessment method can timely detect potential problems, prevent major failures, and improve equipment reliability and production efficiency.

[0112] In step S4, an evaluation is performed according to the degree of abnormality, the abnormal influence range, the abnormal duration, and the abnormal occurrence frequency to obtain an abnormal risk value.

[0113] It should be noted that the abnormal risk value is calculated through the following formula:

[0114] R = w1 A + w 2 S + w 3 D + w 4 F

[0115] Among them, R represents the abnormal risk value, A represents the abnormal degree, S represents the abnormal influence range, D represents the abnormal duration, F represents the occurrence frequency, and w 1 is the weight coefficient of the abnormal degree, w 2 is the weight coefficient of the abnormal influence range, w 3 is the weight coefficient of the abnormal duration, w 4 is the weight coefficient of the occurrence frequency.

[0116] In step S5, adjust according to the abnormal duration of different device types to obtain the abnormal judgment threshold.

[0117] In a specific embodiment, obtain the type information and abnormal duration data of the device. According to the preset abnormal risk level division standard and abnormal judgment threshold, combine the device type and abnormal duration, and dynamically adjust the abnormal risk level division standard and abnormal judgment threshold by using the fuzzy logic algorithm.

[0118] In step S6, judge according to the abnormal risk value and the abnormal judgment threshold; if the abnormal risk value is greater than the abnormal judgment threshold, generate a preliminary alarm signal; among them, the preliminary alarm signal includes the alarm level, alarm reason, and alarm time.

[0119] In a realizable manner, compare the calculated abnormal risk level with the adjusted abnormal judgment threshold. If the abnormal risk level exceeds the abnormal judgment threshold, it is judged that the device is in an abnormal state and an alarm event is triggered. For the triggered alarm event, obtain the preset alarm level division rule, and determine the alarm level of the alarm event according to the abnormal risk level of the device by using the decision tree algorithm. Obtain the cause information that causes the abnormality of the current device, extract keywords, query the association relationship between keywords through the knowledge graph, construct the abnormal cause knowledge graph, and extract the main cause that causes the abnormality from the abnormal cause knowledge graph as the alarm reason. Obtain the current time, associate information such as the alarm level, alarm reason, and current time, and generate structured preliminary alarm signal data.

[0120] In step S7, perform priority arrangement according to the alarm level, the abnormal influence range, and the pre-stored device importance to obtain a hierarchical alarm information queue.

[0121] In the above step S7, the performing priority arrangement according to the alarm level, the abnormal influence range, and the pre-stored device importance to obtain a hierarchical alarm information queue specifically further includes the following steps:

[0122] S71. Calculate the score based on the alarm level, the abnormal influence range, and the pre-stored importance level of the device to obtain the priority score of the alarm event.

[0123] It should be noted that the priority score of the alarm event is calculated through the following formula:

[0124] P = w′ 1 L + w 2 S + w′ 3 I

[0125] Among them, P represents the priority score of the alarm event, L represents the alarm level, S represents the abnormal influence range, I represents the pre-stored importance level of the device, and w' 1 is the weight coefficient of the alarm level, w 2 is the weight coefficient of the abnormal influence range, and w' 3 is the weight coefficient of the pre-stored importance level of the device.

[0126] S72. Sort in descending order according to the priority score of the alarm event to obtain the sequence of alarm events with priority ranking;

[0127] S73. Judge the priority score of the alarm event in the sequence of alarm events with priority ranking; if the priority score of the alarm event is greater than the preset score threshold, it is determined that the alarm event is a high-priority event; otherwise, it is a low-priority event;

[0128] S74. Sort according to the high priority and the low priority of the alarm event to obtain the hierarchical alarm information queue.

[0129] Exemplarily, during the above steps S71 to S74, it is first necessary to extract the key attributes in the preliminary alarm information, such as the alarm level, the affected range, and the importance of the equipment. These attributes together form the basis for hierarchical processing. Taking a petrochemical plant as an example, a temperature anomaly alarm for a reactor contains the following key attributes: the alarm level is high, the affected range involves the entire production line, and the importance of the equipment is a core device. This information will be used as input parameters for the decision tree algorithm. When calculating the priority score of an alarm event, the decision tree algorithm will consider multiple factors. For example, a high alarm level is given a greater weight, and an event with a wide affected range will also receive a higher priority. Suppose the calculated priority score for the temperature anomaly alarm is 85 points (out of 100). This score will be used for subsequent sorting and classification. Priority sorting is a crucial step to ensure that the most urgent problems are handled in a timely manner. In the case of the petrochemical plant, if multiple alarms occur simultaneously, the temperature anomaly alarm will be ranked at the top, while alarms for minor faults of non-critical equipment will be ranked behind. This sorting mechanism can help operators quickly identify and handle the most important problems. The setting of the priority threshold is crucial for distinguishing between high- and low-priority alarms. Suppose the threshold is set at 80 points. Then the temperature anomaly alarm (85 points) will be classified as a high-priority alarm and needs to be handled immediately. This classification method can effectively allocate resources and ensure that key problems receive timely attention. The application of the grouping clustering algorithm can further optimize the alarm handling process. For example, there may be multiple related temperature anomaly alarms. Through clustering, these alarms can be grouped together, facilitating operators to comprehensively understand the problem and adopt a unified solution. This not only improves the processing efficiency but also helps to discover potential systemic problems. Feature selection is an important step in refining alarm information. For a temperature anomaly alarm, key features such as temperature value, pressure value, and flow rate are extracted, while some less relevant parameters are ignored. This feature refinement can reduce data noise and improve the accuracy of subsequent analysis. Finally, the refined alarm information is stored in a queue according to the priority, forming a structured alarm handling sequence. In the example of the petrochemical plant, the temperature anomaly alarm will be placed at the front of the queue to ensure that it can receive the fastest response.

[0130] In step S8, based on the hierarchical alarm information queue, the pre-stored equipment importance, and the abnormal duration, decision analysis is performed to obtain an equipment abnormal alarm signal.

[0131] In the above step S8, the decision analysis based on the hierarchical alarm information queue, the pre-stored equipment importance, and the abnormal duration to obtain an equipment abnormal alarm signal specifically further includes the following steps:

[0132] S81, using a preset alarm decision algorithm to perform multi-level factor analysis on the hierarchical alarm information queue to obtain an alarm information sending method;

[0133] S82. Select the recipients according to the pre-stored importance level of the device to obtain the recipients of the alarm information.

[0134] S83. Set according to the abnormal duration to obtain the alarm information sending frequency.

[0135] S84. Combine according to the alarm information sending method, the recipients of the alarm information and the alarm information sending frequency to obtain the device abnormal alarm signal.

[0136] In a specific embodiment, in the above steps S81 to S84, the specific implementation process includes: extracting attributes such as the alarm level, the influence range, the duration, and the device type according to the alarm information in the hierarchical alarm information queue. According to the preset alarm decision algorithm, comprehensively consider attributes such as the alarm level, the influence range, the duration, and the device type, and calculate the priority score of the alarm information. According to the priority score of the alarm information, use the decision tree algorithm to divide the alarm information into different alarm levels to obtain the alarm information sending method matching the alarm level. Through the device asset management system, obtain the importance level attribute of the device, and according to the importance level attribute of the device, use the association rule mining algorithm to determine the recipients of the alarm information corresponding to the devices with different importance levels. For the duration attribute of the alarm information, use the time series analysis algorithm to judge whether the abnormal duration exceeds the preset threshold. If it exceeds the threshold, dynamically adjust the alarm information sending frequency according to the length of the abnormal duration. Combine attributes such as the alarm level, the alarm information sending method, the recipients of the alarm information, and the alarm information sending frequency, and generate a device abnormal alarm signal matching it through the rule engine technology.

[0137] Exemplarily, the alarm information processing system first obtains alarm information from the hierarchical alarm information queue and extracts key attributes such as alarm level, impact scope, duration, and device type. These attributes provide basic data for subsequent decision-making. For example, a temperature sensor in a chemical plant detects an abnormal increase in the temperature of a reactor. The system extracts the alarm level as "urgent", the impact scope as "the entire production line", the duration as "30 minutes", and the device type as "critical production equipment". The preset alarm decision algorithm comprehensively considers these attributes and calculates the priority score of the alarm information. In the above example, since the abnormal temperature may lead to serious safety accidents, the system will give a relatively high priority score. The decision tree algorithm classifies the alarm information into different levels, such as "urgent", "important", "general", etc., according to this score and matches the corresponding sending methods. For the abnormal temperature of the reactor, the system will classify it as the "urgent" level and select multiple notification methods such as phone calls, text messages, and application push notifications. The equipment asset management system provides information on the importance of the equipment, and the association rule mining algorithm determines the recipients of the alarm information based on this. In the case of the chemical plant, as the reactor is a core production equipment, its abnormal situation will be notified to the plant supervisor, safety officer, and relevant operators simultaneously. This method ensures that critical information can be conveyed to the appropriate handling personnel in a timely manner. The time series analysis algorithm is used to determine whether the abnormal duration exceeds a preset threshold and adjusts the sending frequency of the alarm information accordingly. If the temperature of the reactor continues to be abnormal for more than the preset 30-minute threshold, the system will increase the sending frequency of the alarm information, from once every 30 minutes initially to once every 10 minutes, to ensure that relevant personnel continuously pay attention to the problem. Finally, the rule engine technology combines attributes such as alarm level, sending method, recipient, and sending frequency to generate a matching device abnormal alarm signal. In the case of the abnormal temperature of the reactor, the generated alarm signal may contain the following information: urgent level, phone call and text message notification, sent to the plant supervisor and safety officer, and sent once every 10 minutes. This comprehensive consideration method can ensure the timely and accurate transmission of alarm information and improve the processing efficiency of abnormal situations. Through this multi-dimensional alarm information processing method, the system can intelligently analyze and respond to various device abnormal situations, greatly improving the safety and efficiency of industrial production. It can not only quickly identify and classify problems, but also dynamically adjust response strategies according to the severity and duration of the problems, ensuring that critical information can be conveyed to the appropriate handling personnel in a timely manner, thereby minimizing potential losses and risks.

[0138] In summary, the present invention discloses a micro-monitoring and alarm method based on multi-modal recognition, including obtaining real-time monitoring data; performing feature analysis on the real-time monitoring data to obtain a state feature vector; performing anomaly analysis on the state feature vector to obtain the degree of anomaly, the scope of anomaly influence, the duration of anomaly, and the frequency of anomaly occurrence; evaluating according to the degree of anomaly, the scope of anomaly influence, the duration of anomaly, and the frequency of anomaly occurrence to obtain an anomaly risk value; adjusting according to the duration of anomaly of different device types to obtain an anomaly judgment threshold; judging according to the anomaly risk value and the anomaly judgment threshold; if the anomaly risk value is greater than the anomaly judgment threshold, generating a preliminary alarm signal; wherein, the preliminary alarm signal includes an alarm level, an alarm reason, and an alarm time; performing priority ranking according to the alarm level, the scope of anomaly influence, and the importance of the device stored in advance to obtain a hierarchical alarm information queue; performing decision-making analysis according to the hierarchical alarm information queue, the importance of the device stored in advance, and the duration of anomaly to obtain a device anomaly alarm signal. By collecting multi-source heterogeneous data such as vibration, temperature, pressure, and sound of the device and performing real-time fusion analysis, the present invention generates a feature vector reflecting the comprehensive operating state of the device; combined with a knowledge base and intelligent algorithms, the present invention can accurately judge the degree of anomaly and potential risks of the device, and dynamically adjust the alarm threshold according to factors such as device type and duration of anomaly; through a multi-level alarm processing mechanism, the present invention can rank the priority of anomaly events, and intelligently select alarm methods and recipients according to the alarm level, influence scope, etc. The present invention significantly improves the real-time performance and accuracy of device monitoring, realizes early warning and accurate alarm of anomalies, effectively reduces the risk of device failures, and improves the safety and efficiency of industrial production.

[0139] Referring to Figure 2 , the second embodiment of the present invention provides a micro-monitoring and alarm system based on multi-modal recognition, including:

[0140] A data acquisition module 101 for obtaining real-time monitoring data;

[0141] A feature analysis module 102 for performing feature analysis on the real-time monitoring data to obtain a state feature vector;

[0142] An anomaly analysis module 103 for performing anomaly analysis on the state feature vector to obtain the degree of anomaly, the scope of anomaly influence, the duration of anomaly, and the frequency of anomaly occurrence;

[0143] A risk assessment module 104 for evaluating according to the degree of anomaly, the scope of anomaly influence, the duration of anomaly, and the frequency of anomaly occurrence to obtain an anomaly risk value;

[0144] A threshold determination module 105, configured to adjust according to the abnormal duration of different device types to obtain an abnormal judgment threshold;

[0145] An abnormal judgment module 106, configured to judge according to the abnormal risk level and the abnormal judgment threshold; if the abnormal risk value is greater than the abnormal judgment threshold, a preliminary alarm signal is generated; wherein, the preliminary alarm signal includes an alarm level, an alarm reason, and an alarm time;

[0146] An alarm sorting module 107, configured to perform priority sorting according to the alarm level, the abnormal influence range, and the importance of the device stored in advance to obtain a hierarchical alarm information queue;

[0147] A signal generation module 108, configured to perform decision analysis according to the hierarchical alarm information queue, the importance of the device stored in advance, and the abnormal duration to obtain a device abnormal alarm signal.

[0148] It should be noted that a micro-monitoring alarm system based on multi-modal recognition provided in an embodiment of the present invention is used to execute all the process steps of a micro-monitoring alarm method based on multi-modal recognition in the above embodiment. The working principles and beneficial effects of the two correspond one by one, so they will not be elaborated here.

[0149] An embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a micro-monitoring alarm program based on multi-modal recognition. When the processor executes the computer program, the steps in the above embodiments of the micro-monitoring alarm method based on multi-modal recognition are implemented, such as Figure 1 Step S1 shown. Alternatively, when the processor executes the computer program, the functions of each module / unit in the above device embodiments are implemented, such as a data acquisition module.

[0150] Exemplarily, the computer program may be divided into one or more modules / units. The one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0151] The electronic device may be a computing device such as a desktop computer, notebook, handheld computer, and smart tablet, etc. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above components are only examples of the electronic device and do not constitute a limitation on the electronic device. It may include more or fewer components than the above, or combine some components, or different components. For example, the electronic device may also include input / output devices, network access devices, a bus, etc.

[0152] The so-called processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the electronic device and connects various parts of the entire electronic device through various interfaces and circuits.

[0153] The memory may be used to store the computer programs and / or modules. The processor realizes various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory, and by calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. Among them, the program storage area may store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area may store data created according to the use of the mobile phone (such as audio data, phone book, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one magnetic disk storage device, flash device, or other volatile solid-state storage devices.

[0154] Among them, if the modules / units integrated in the electronic device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-described embodiment methods of the present invention, it can also be completed by a computer program instructing relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0155] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the attached drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that they have a communication connection, which can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0156] The above-described specific embodiments have further elaborated on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only for the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. It is particularly pointed out that for those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A micro-monitoring alarm method based on multimodal recognition, characterized in that: Executed by a computer, including: Obtain real-time monitoring data; Performing feature analysis based on the real-time monitoring data to obtain a state feature vector; Performing anomaly analysis according to the state feature vector to obtain the degree of anomaly, the impact range of anomaly, the duration of anomaly and the frequency of anomaly occurrence; An abnormality risk value is obtained by evaluating the abnormality degree, the abnormality impact range, the abnormality duration and the abnormality occurrence frequency; Adjusting the abnormality duration according to different equipment types to obtain an abnormality judgment threshold; A judgment is made according to the abnormal risk value and the abnormal judgment threshold; if the abnormal risk value is greater than the abnormal judgment threshold, a preliminary alarm signal is generated; wherein the preliminary alarm signal includes an alarm level, an alarm reason and an alarm time; Prioritize according to the alarm level, the abnormal impact range and the importance of the pre-stored equipment to obtain a hierarchical alarm information queue; A decision analysis is performed based on the hierarchical alarm information queue, the pre-stored equipment importance and the abnormality duration to obtain an equipment abnormality alarm signal.

2. The micro-monitoring alarm method based on multimodal recognition according to claim 1 is characterized in that: The real-time monitoring data refers to data obtained by monitoring using different collection strategies based on the health status assessment results of the equipment.

3. The micro-monitoring alarm method based on multimodal recognition according to claim 1 is characterized in that: The abnormality analysis is performed according to the state feature vector to obtain the abnormality degree, abnormal impact range, abnormality duration and abnormality occurrence frequency, including: Analyzing the state feature vector using a knowledge-based intelligent analysis algorithm to obtain an abnormality assessment function; Calculate the degree of membership according to the abnormality evaluation function to obtain the degree of abnormality; A judgment is made based on the abnormality level and a preset abnormality level threshold; if the abnormality level is greater than the preset abnormality level threshold, it is determined that an abnormality has occurred in the device, and the impact range, duration and frequency of the abnormality are counted to obtain the abnormal impact range, abnormal duration and abnormal occurrence frequency.

4. The micro-monitoring alarm method based on multimodal recognition according to claim 3 is characterized in that: The analyzing the state feature vector using a knowledge-based intelligent analysis algorithm to obtain an abnormality assessment function includes: The anomaly assessment function is calculated by the following formula: Among them, f i (x i ) represents the abnormal evaluation function of the i-th state feature vector, x i is the current value of the eigenvector of the i-th state, μ i is the normal mean of the eigenvector of the i-th state, σ i is the standard deviation of the eigenvector of the i-th state.

5. The micro-monitoring alarm method based on multimodal recognition according to claim 1 is characterized in that: The abnormality risk value is obtained by evaluating the abnormality degree, the abnormality impact range, the abnormality duration and the abnormality occurrence frequency, including: The abnormal risk value is calculated by the following formula: R=w1A+w2S+w3D+w4F Among them, R represents the abnormal risk value, A represents the abnormal degree, S represents the abnormal impact range, D represents the abnormal duration, F represents the occurrence frequency, w1 is the abnormal degree weight coefficient, w2 is the abnormal impact range weight coefficient, w3 is the abnormal duration weight coefficient, and w4 is the occurrence frequency weight coefficient.

6. The micro-monitoring alarm method based on multimodal recognition according to claim 1 is characterized in that: The priority arrangement is performed according to the alarm level, the abnormal impact range and the importance of the pre-stored equipment to obtain a hierarchical alarm information queue, including: Calculate the score according to the alarm level, the impact range of the abnormality and the importance of the equipment stored in advance to obtain the alarm event priority score; Sorting the alarm events in descending order according to their priority scores to obtain a priority-sorted alarm event sequence; Determining the priority score of the alarm event in the priority-ranked alarm event sequence; if the priority score of the alarm event is greater than a preset score threshold, determining that the alarm event is a high priority; otherwise, it is a low priority; The alarm events are sorted according to the high priority and the low priority to obtain a hierarchical alarm information queue.

7. The micro-monitoring alarm method based on multimodal recognition according to claim 6 is characterized in that: The score calculation is performed according to the alarm level, the abnormal impact range and the pre-stored equipment importance to obtain the alarm event priority score, including: The alarm event priority score is calculated using the following formula: P=w′1L+w2S+w′3I Among them, P represents the alarm event priority score, L represents the alarm level, S represents the abnormal impact range, I represents the pre-stored equipment importance, w'1 is the alarm level weight coefficient, w2 is the abnormal impact range weight coefficient, and w'3 is the pre-stored equipment importance weight coefficient.

8. The micro-monitoring alarm method based on multimodal recognition according to claim 1 is characterized in that: The decision analysis is performed according to the hierarchical alarm information queue, the pre-stored equipment importance and the abnormality duration to obtain the equipment abnormality alarm signal, including: A preset alarm decision algorithm is used to perform multi-level factor analysis according to the hierarchical alarm information queue to obtain an alarm information sending method; Selecting a recipient according to the importance of the pre-stored device to obtain an alarm information recipient; According to the abnormal duration, the alarm information sending frequency is set; The device abnormality alarm signal is obtained by combining the alarm information sending method, the alarm information receiving object and the alarm information sending frequency.

9. A micro-monitoring alarm system based on multimodal recognition, characterized in that: include: A data acquisition module, used to acquire real-time monitoring data; A feature analysis module, used to perform feature analysis based on the real-time monitoring data to obtain a state feature vector; An abnormality analysis module, used to perform abnormality analysis according to the state feature vector to obtain the abnormality degree, abnormal impact range, abnormality duration and abnormality occurrence frequency; A risk assessment module, used to evaluate the abnormality degree, the abnormality impact range, the abnormality duration and the abnormality occurrence frequency to obtain an abnormality risk value; A threshold determination module, used to adjust the abnormality duration according to different equipment types to obtain an abnormality judgment threshold; An abnormality judgment module, used to make a judgment according to the abnormality risk level and the abnormality judgment threshold; if the abnormality risk value is greater than the abnormality judgment threshold, a preliminary alarm signal is generated; wherein the preliminary alarm signal includes an alarm level, an alarm reason and an alarm time; An alarm sorting module is used to prioritize the alarms according to the alarm level, the impact range of the abnormality and the importance of the pre-stored equipment to obtain a hierarchical alarm information queue; The signal generation module is used to perform decision analysis based on the hierarchical alarm information queue, the pre-stored equipment importance and the abnormality duration to obtain an equipment abnormality alarm signal.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the micro-monitoring alarm method based on multimodal recognition as described in any one of claims 1 to 8.

Images